Category: IPD Publication

A diffusion model for protein design

A team led by Baker Lab scientists Joseph Watson, David Juergens, Nate Bennett, Brian Trippe, and Jason Yim has created a powerful new way to design proteins by combining structure prediction networks and generative diffusion models. The team demonstrated extremely high computational success and tested hundreds of A.I.-generated proteins in the lab, finding that many may be useful as medications, vaccines, or even new nanomaterials. This research will soon be available as a preprint on bioRvix titled “Broadly applicable and accurate protein design by integrating structure prediction networks and diffusion generative models.” An advanced copy is available here. A similar preprint by Generate Biomedicines is also available here.

The software tool DALL-E produces high-quality images that have never existed before using something called a diffusion model, which is a machine-learning algorithm that specializes in adding and removing noise. Diffusion models for image generation begin with grainy bits of static and gradually remove noise until a clear picture is formed. Additional pieces of software guide this de-noising process so that the new images end up matching what was asked for.

We have developed a guided diffusion model for generating new proteins. With prior design methods, tens of thousands of molecules may have to be tested before finding a single one that performs as intended. Using the new design method, dubbed RF diffusion, the team had to test as little as one per design challenge. RF diffusion outperforms existing protein design methods across a broad range of problems. Highlights include a picomolar binder generated through pure computation and a series of novel symmetric assemblies experimentally confirmed by electron microscopy.

“These works reveal just how powerful diffusion models can be for protein design,” says Watson. “It’s extremely exciting,” added Juergens, “and it’s really just the beginning.”

RF diffusion can generate novel proteins that bind to molecular targets


RF diffusion can be configured to produce symmetric or asymmetric oligomers.

ProteinMPNN excels at creating new proteins

Over the past two years, machine learning has revolutionized protein structure prediction. Now, three papers in Science describe a similar revolution in protein design. In the new papers, scientists in the Baker lab show that machine learning can be used to create proteins much more accurately and quickly than previously possible. This could lead to many new vaccines, treatments, tools for carbon capture, and sustainable biomaterials.

“Proteins are fundamental across biology, but we know that all the proteins found in every plant, animal, and microbe make up far less than one percent of what is possible. With these new software tools, should be able to find solutions to long-standing challenges in medicine, energy, and technology,” said senior author David Baker.

To go beyond the proteins found in nature, our team broke down the challenge of protein design into three parts and used new software solutions, including ProteinMPNN, for each. 

First, a new protein shape must be generated. In a paper published on July 21 in the journal Science, we showed that artificial intelligence can create new proteins that may be useful as vaccines, cancer treatments, or even tools for pulling carbon pollution out of the air. The team developed two strategies for designing new protein structures. The first, dubbed “hallucination,” is akin to DALL-E or other generative A.I. tools that produce output based on simple prompts. The second, dubbed “inpainting,” is analogous to the autocomplete feature found in modern search bars. “Most people can come up with new images of cats or write a paragraph from a prompt if asked, but with protein design, the human brain cannot do what computers now can,” said project scientist Jue Wang.

Second, to speed up the process, the team led by Justas Dauparas from the Baker lab devised a new algorithm for generating amino acid sequences. Described in the September 15 issue of Science, this software tool — called ProteinMPNN — runs in about one second, which is more than 200 times faster than the previous best software. Its results are superior to prior tools, and the software requires no expert customization to run. “Neural networks are easy to train if you have a ton of data, but with proteins, we don’t have as many examples as we would like. We had to go in and identify which features in these molecules are the most important. It was a bit of trial and error,” said project scientist Justas Dauparas.

Artificial-intelligence tools are helping to scientists to come up with proteins that are shaped unlike anything in nature.
Nature: Scientists are using AI to dream up revolutionary new proteins

Third, we used AlphaFold, a tool developed by Alphabet’s DeepMind, to independently assess whether our designed amino acid sequences were likely to fold into the intended shapes. “Software for predicting protein structures is part of the solution but it cannot come up with anything new on its own,” explained Dauparas. “Even if you had a perfect tool for predicting how protein sequences fold, you would have to search through billions and billions of sequences to find any new useful proteins.”

“ProteinMPNN is to protein design what AlphaFold was to protein structure prediction,” said Baker.

In another paper appearing in Science, a team led by Basile Wicky, Lukas Milles, and Alexis Courbet from the Baker lab confirmed that ProteinMPNN together with the other new machine learning tools could reliably generate proteins that functioned in the laboratory. “It’s not enough to trust that the computer is designing proteins well — you have to actually study these molecules in the real world. We found that proteins made using ProteinMPNN were much more likely to fold up as intended, and we could create very complex protein assemblies using these methods” said project scientist Basile Wicky.

MIT Tech Review: An AI that can design new proteins could help unlock new cures and materials

Among the new proteins made were nanoscale rings that the researchers believe could be used as parts for custom nanomachines. Electron microscopes were used to observe the rings, which have diameters roughly a billion times smaller than a poppy seed.

“This is the very beginning of machine learning in protein design. In the coming months, we will be working to improve these tools to create even more dynamic and functional proteins,” said Baker. 


Compute resources for this work were donated by Microsoft and Amazon Web Services. Funding was provided by the Audacious Project at the Institute for Protein Design; Microsoft; Eric and Wendy Schmidt by recommendation of the Schmidt Futures; the DARPA Synergistic Discovery and Design project (HR001117S0003 contract FA8750-17-C-0219); the DARPA Harnessing Enzymatic Activity for Lifesaving Remedies project (HR001120S0052 contract HR0011-21-2-0012); the Washington Research Foundation; the Open Philanthropy Project Improving Protein Design Fund; Amgen; the Alfred P. Sloan Foundation Matter-to-Life Program Grant (G-2021-16899); the Donald and Jo Anne Petersen Endowment for Accelerating Advancements in Alzheimer’s Disease Research; the Human Frontier Science Program Cross Disciplinary Fellowship (LT000395/2020-C); the European Molecular Biology Organization (ALTF 139-2018), including an EMBO Non-Stipendiary Fellowship (ALTF 1047-2019) and an EMBO Long-term Fellowship (ALTF 191-2021); the “la Caixa” Foundation; the Howard Hughes Medical Institute, including a Hanna Gray fellowship (GT11817); the National Science Foundation (MCB 2032259, CHE-1629214, DBI 1937533, DGE-2140004); the National Institutes for Health (DP5OD026389); the National Institute of Allergy and Infectious Diseases (HHSN272201700059C); the National Institute on Aging (5U19AG065156); the National Institute of General Medical Sciences (P30 GM124169-01, P41 GM 103533-24); the National Cancer Institute (R01CA240339); the Swiss National Science Foundation; the Swiss National Center of Competence for Molecular Systems Engineering; the Swiss National Center of Competence in Chemical Biology; and the European Research Council (716058).

Design of membrane-traversing peptides leads to new spinout

Researchers at the Institute for Protein Design have discovered how to create peptides that slip through membranes and enter cells. This drug design breakthrough may lead to new medications for a wide variety of health disorders, including cancer, infection, and inflammation. This research appears in the journal Cell [PDF].

“This new ability to design membrane-permeable peptides with high structural accuracy opens the door to a new class of medicines that combine the advantages of traditional small-molecule drugs and larger protein therapeutics,” said senior author David Baker, director of the Institute.

Peptide drugs are made from the same building blocks as proteins. Unlike many traditional small molecule medicines, peptides can bind protein targets in the body with great precision, promising fewer side effects. 

Gaurav Bhardwaj, Adam Moyer, Naozumi Hiranuma, Patrick Salveson, and David Baker at the UW Medicine Institute for Protein Design have co-founded a new company, Vilya, Inc., with ARCH Venture Partners that is licensing the platform and molecules described in the paper.

“We know that peptides can be excellent medicines, but a big problem is that they don’t get into cells. There are a lot of great drug targets inside our cells, and if we can get in there, that space opens up.”

Lead author and IPD faculty member Gaurav Bhardwaj, assistant professor of medicinal chemistry

 Discovering how to send peptides through membranes was a chemical problem. “Membranes are made of lipids, and most peptides have chemical features that cause them to hold onto water molecules. Dragging these water molecules through the lipids is difficult,” explains Bhardwaj. The scientists tried several solutions. They first crafted peptides with chemical features that reduce interactions with water. In another approach, they designed peptides that could change shapes while crossing membranes.

To evaluate their new peptide design strategies, the team synthesized over 180 custom molecules. Laboratory tests on artificial membranes revealed that most of the peptides could pass through lipids as desired. Further tests involving gut epithelial cells led the scientists to believe that some of the new peptides could make the jump from the stomach into the bloodstream. Animal studies confirmed that when swallowed, some of the peptides could efficiently move out of the gut, cross several membranes, and enter living cells.

The Core Research Labs at the Institute for Protein Design determined high-resolution structures for 36 of the new compounds, confirming that the peptides adopted the precise shapes that the designers intended.

“These molecules are promising starting points for future drugs. My lab is now working to turn them into antibiotics, antivirals, and cancer treatments,” said Bhardwaj.

Bhardwaj believes it may now be possible to create peptide drugs that treat COVID-19. “One of the most obvious drug targets is located inside infected cells. If we could shut down that enzyme, that would prevent the virus from creating more copies of itself.”

Antibiotics are another area of focus. “Antibiotic resistance is becoming a problem for most drugs, and the discovery of new antibiotics from nature has been slow,” explains Bhardwaj. “We are using rational design to try to create new peptide-based antibiotics. Here, the fact that these molecules can be swallowed rather than injected is going to be very important for their effective use and broader adoption ”

Scientists at the University of Washington developed the design methods, created the new peptides, and performed some laboratory tests. Permeability using gut epithelial cells was measured in the David Craik lab at the University of Queensland. Structural characterization of the new peptides was performed by the Gaetano Montelione lab at Rensselaer Polytechnic Institute.

The paper, “Accurate de novo design of membrane-traversing macrocycles,” appears online in Cell on August 29. The research was supported by The Audacious Project; Gates Ventures; Eric and Wendy Schmidt by recommendation of the Schmidt Futures; the Nordstrom Barrier Institute for Protein Design Directors Fund; Wu Tsai Translational Fund; Bill and Melinda Gates Foundation (INV-010680; Open Philanthropy Project; Takeda Pharmaceuticals; Howard Hughes Medical Institute; Washington State Supplement Funding; Department of Defense; Simons Foundation; Defense Threat Reduction Agency (HDTRA1-19-1-0003); National Institutes of Health (R01AG063845, R01GM120574, R35-GM141818, F32GM120791-02); and Washington Research Foundation.

Vilya Launches to Design a New Transformational Class of Medicines That Precisely Target Disease Biology (

Training A.I. to generate medicines and vaccines

Today we report in Science [PDF] the development of artificial intelligence software that can create proteins that may be useful as vaccines, cancer treatments, or even tools for pulling carbon pollution out of the air.

This project was led by Jue Wang, Doug Tischer, and Joseph L. Watson, who are postdoctoral scholars at UW Medicine, as well as Sidney Lisanza and David Juergens, who are graduate students at UW Medicine. Senior authors include Sergey Ovchinnikov, a John Harvard Distinguished Science Fellow at Harvard University, and David Baker, professor of biochemistry, HHMI Investigator, and director of the Institute for Protein Design at UW Medicine.

“The proteins we find in nature are amazing molecules, but designed proteins can do so much more,” said Baker. “In this work, we show that machine learning can be used to design proteins with a wide variety of functions.”

Training new neural networks

Inspired by how machine learning algorithms can generate stories or even images from prompts, the team set out to build similar software for designing new proteins. “The idea is the same: neural networks can be trained to see patterns in data. Once trained, you can give it a prompt and see if it can generate an elegant solution. Often the results are compelling — or even beautiful,” said lead author Joseph Watson.

The team trained multiple neural networks using information from the Protein Data Bank, which is a public repository of hundreds of thousands of protein structures from across all kingdoms of life. The neural networks that resulted have surprised even the scientists who created them.

The team developed two approaches for designing proteins with new functions. The first, dubbed “hallucination” is akin to DALL-E or other generative A.I. tools that produce new output based on simple prompts. The second, dubbed “inpainting,” is analogous to the autocomplete feature found in modern search bars and email clients.

“Most people can come up with new images of cats or write a paragraph from a prompt if asked, but with protein design, the human brain cannot do what computers now can,” said lead author Jue Wang. “Humans just cannot imagine what the solution might look like, but we have set up machines that do.”

Starting with gibberish

To explain how the neural networks ‘hallucinate’ a new protein, the team compares it to how it might write a book: “You start with a random assortment of words — total gibberish. Then you impose a requirement such as that in the opening paragraph, it needs to be a dark and stormy night. Then the computer will change the words one at a time and ask itself ‘Does this make my story make more sense?’ If it does, it keeps the changes until a complete story is written,” explains Wang.

Both books and proteins can be understood as long sequences of letters. In the case of proteins, each letter corresponds to a chemical building block called an amino acid. Beginning with a random chain of amino acids, the software mutates the sequence over and over until a final sequence that encodes the desired function is generated. These final amino acid sequences encode proteins that can then be manufactured and studied in the laboratory.

Autocomplete for proteins

The team also showed that neural networks can fill in missing pieces of a protein structure in only a few seconds. Such software could aid in the development of new medicines.

“With autocomplete, or “protein Inpainting”, we start with the key features we want to see in a new protein, then let the software come up with the rest. Those features can be known binding motifs or even enzyme active sites,” explains Watson. Laboratory testing revealed that many proteins generated through hallucination and inpainting functioned as intended. This included novel proteins that can bind metals as well as those that bind the anti-cancer receptor PD-1.

Creating new vaccines

The new neural networks can generate several different kinds of proteins in as little as one second. Some include potential vaccines for the deadly respiratory virus RSV.

All vaccines work by presenting a piece of a pathogen to the immune system. Scientists often know which piece would work best, but creating a vaccine that achieves a desired molecular shape can be challenging. Using the new neural networks, the team prompted a computer to create new proteins that included the necessary pathogen fragment as part of their final structure. The software was free to create any supporting structures around the key fragment, yielding several potential vaccines with diverse molecular shapes.

When tested in the lab, the team found that known antibodies against RSV stuck to three of their hallucinated proteins. This confirms that the new proteins adopted their intended shapes and suggests they may be viable vaccine candidates that could prompt the body to generate its own highly specific antibodies. Additional testing, including in animals, is still needed.

“I started working on the vaccine stuff just as a way to test our new methods, but in the middle of working on the project, my two-year-old son got infected by RSV and spent an evening in the ER to have his lungs cleared. It made me realize that even the ‘test’ problems we were working on were actually quite meaningful,” said Wang.

“These are very powerful new approaches, but there is still much room for improvement,” said Baker, who was a recipient of the 2021 Breakthrough Prize in Life Sciences. “Designing high activity enzymes, for example, is still very challenging. But every month our methods just keep getting better! Deep learning transformed protein structure prediction in the past two years, we are now in the midst of a similar transformation of protein design.”


Compute resources for this work were donated by Microsoft and Amazon Web Services. Funding was provided by the Audacious Project at the Institute for Protein Design; Microsoft; Eric and Wendy Schmidt by recommendation of the Schmidt Futures; the DARPA Synergistic Discovery and Design project (HR001117S0003 contract FA8750-17-C-0219); the DARPA Harnessing Enzymatic Activity for Lifesaving Remedies project (HR001120S0052 contract HR0011-21-2-0012); the Washington Research Foundation; the Open Philanthropy Project Improving Protein Design Fund; Amgen; the Human Frontier Science Program Cross Disciplinary Fellowship (LT000395/2020-C) and EMBO Non-Stipendiary Fellowship (ALTF 1047-2019); the EMBO Fellowship (ALTF 191-2021); the European Molecular Biology Organization (ALTF 139-2018); the “la Caixa” Foundation; the National Institute of Allergy and Infectious Diseases (HHSN272201700059C), the National Institutes for Health (DP5OD026389); the National Science Foundation (MCB 2032259); the Howard Hughes Medical Institute, the National Institute on Aging (5U19AG065156); the National Cancer Institute (R01CA240339); the Swiss National Science Foundation; the Swiss National Center of Competence for Molecular Systems Engineering; the Swiss National Center of Competence in Chemical Biology; and the European Research Council (716058).

Custom biosensors for detecting coronavirus antibodies in blood

Today we report in Nature Biotechnology the design of custom protein-based biosensors that can detect coronavirus-neutralizing antibodies in blood. This research, which builds on prior sensor design technology developed in the Baker lab, was led by Baker lab postdoctoral scholars Jason Zhang, PhD, and Hsien-Wei (Andy) Yeh, PhD.

From Behind the Paper:

[W]e utilized the de novo designed LOCKR (Latching, Orthogonal Cage/Key pRotein) system as a biosensor for measuring SARS-CoV-2 components and antibodies. The two-state LOCKR system is designed to be switchable, thus ideal for use as a biosensor2. LOCKR contains 2 proteins: 1) Cage protein: contains a 5-helical cage domain tethered to and interacting with the 1-helical latch domain, 2): Key protein: contains the 1-helical key domain that also has affinity to the cage domain. To transform LOCKR into a sensor for SARS-CoV-2 components (specifically the receptor binding domain (RBD) from the spike protein), a de novo designed binder (with picomolar affinity) to RBD called LCB13 was embedded on the end of the latch so that binding of RBD to LCB1 weakens the binding between cage and latch domains, strengthening the binding between cage and key domains, and thus allowing for the 2 LOCKR proteins to associate. To allow for readout of this binding event, split luciferase was added to the LOCKR proteins where the smaller bit was embedded in the latch and the larger portion attached to the end of the key protein. For this RBD sensor, the cage protein is called lucCageRBD and the key protein is called lucKey2. Thus, increased amounts of RBD binding to LCB1 in the cage protein translates to increased bioluminescence from the now reconstituted luciferase.


Rotory proteins designed from scratch

Today we report in Science the design of rotary devices made from custom proteins. These microscopic “axles” and “rotors” come together to form spinning assemblies, rather than being locked in just one orientation. Such mechanical coupling is a key feature of any machine.

The new axle-rotor devices — which are each about a billion times smaller than a poppy seed — were designed on computers, produced inside living cells, and studied in the lab.

This research paves the way for a new generation of nanoscale machines in which the motion of the components is powered by solar energy or chemical fuel.

The project was led by biochemist Alexis Courbet, a postdoctoral scholar in Baker lab, and by Jesse Hansen, a recent graduate student in the laboratory of Justin Kohlman, an associate professor of biochemistry at UW Medicine.

Alexis Courbet, PhD

“One of our goals is to create nanomachines that might one day circulate through the blood and autonomously remove unwanted plaques or even cancer cells,” said Courbet. “We know that very complex machines can be assembled from simple parts.”

Scientists from the Kohlman and Veesler labs at UW Medicine used electron microscopes to visualize the rotation of the new protein devices.

This work was supported in part by The Audacious Project at the Institute for Protein Design, Open Philanthropy, National Science Foundation, and a Washington Research Foundation Senior Fellowship. A full list of funders can be found in the manuscript.

Protein drugs designed from the ground up

Today we report in Nature a new method for generating protein drugs. Using Rosetta-based design, an international team designed molecules that can target important proteins in the body, such as the insulin receptor, as well as proteins on the surface of viruses. This solves a long-standing challenge in drug development and may lead to new treatments for cancer, diabetes, infection, inflammation, and beyond.

This work was led by two Baker lab postdoctoral scholars – Longxing Cao, PhD, and Brian Coventry, PhD – who combined recent advances in computational protein design to arrive at a strategy for creating new proteins that bind molecular targets in a manner similar to antibodies. They developed software that can scan a target molecule, identify potential binding sites, generate proteins targeting those sites, and then screen from millions of candidate binding proteins to identify those most likely to function.

The team generated high-affinity binding proteins against 12 distinct molecular targets. These targets include important cellular receptors such as TrkA, EGFR, Tie2, and the insulin receptor, as well proteins on the surface of the influenza virus and SARS-CoV-2 (the virus that causes COVID-19).

“When it comes to creating new drugs, there are easy targets and there are hard targets,” said Cao, who is now an assistant professor at Westlake University. “In this paper, we show that even very hard targets are amenable to this approach. We were able to make binding proteins to some targets that had no known binding partners or antibodies.”

In total, the team produced over half a million candidate binding proteins for the 12 selected molecular targets. Data collected on this large pool of candidate binding proteins was used to improve the overall method.

“We look forward to seeing how these molecules might be used in a clinical context, and more importantly how this new method of designing protein drugs might lead to even more promising compounds in the future,” said Coventry.

The research team included scientists from the University of Washington School of Medicine, Yale University School of Medicine, Stanford University School of Medicine, Ghent University, The Scripps Research Institute, and the National Cancer Institute, among other institutions.

This work was supported in part by The Audacious Project at the Institute for Protein Design, Open Philanthropy Project, National Institutes of Health (HHSN272201700059C, R01AI140245, R01AI150855, R01AG063845), Defense Advanced Research Project Agency (HR0011835403 contract FA8750-17-C-0219), Defense Threat Reduction Agency (HDTRA1-16-C-0029), Schmidt Futures, Gates Ventures, Donald and Jo Anne Petersen Endowment, and an Azure computing gift for COVID-19 research provided by Microsoft.

Diverse protein assemblies by (negative) design

A new approach for creating custom protein complexes yields asymmetric assemblies with interchangeable parts.

Today we report in Science the design of new protein assemblies made from modular parts. These complexes — which adopt linear, branching, or closed-loop architectures — contain up to six unique proteins, each of which remains folded and soluble in the absence of any binding partners. Baker lab postdoctoral scholars Danny Sahtoe and Florian Praetorius led this work.

Proteins in living cells often come together to form complexes that carry out vital functions. A key feature of these sophisticated assemblies is their ability to exchange parts as needed. To replicate DNA, for example, dozens of unique proteins in a cell spontaneously form into clamps, clamp-loading assemblies, and multi-chain enzyme complexes.

“We see many areas in synthetic biology that would benefit from the ability to pair up proteins in structurally defined ways, but there is currently no easy way to do this,” said Praetorius. “For this project, we wanted to create stable proteins that could spontaneously and rapidly assemble into predictable configurations upon mixing. These could then be fused to other molecules to drive them together.”

Lead authors Danny Sahtoe, PhD, and Florian Praetorius, PhD

Implicit negative design

To begin, the team first sought to create new pairs of proteins that would bind reliably to their cognate partners but not themselves. Such stable heterodimers could then serve as junction points for larger multi-protein architectures.

“To create stable heterodimers, we turned to negative design,” explains Sahtoe. “We decided to introduce three properties into our proteins that make self-association unlikely. First, we made our proteins stable by including real hydrophobic cores. Second, we designed interfaces that require two beta-sheets to pair up, creating unique and continuous beta-sheets across the heterodimer interfaces. And finally, we used Rosetta to model the possible unwanted homooligomeric states and then used that information to prevent those states.”

Using these design principles, the team generated several new heterodimeric proteins that assemble rapidly, even in crowded cell lysates. High-resolution crystal structures of three such dimers confirmed the intended binding modes.

These well-behaved proteins were then recombined and fused to create new “connector proteins” that could bind partners at either end. The team went on to create branched, star-shaped, and angled connectors as well. These modular parts can in principle be mixed and matched to create roughly 500 unique multi-chain assemblies.

The new self-assembling protein parts function as designed in living cells, mediating the assembly of liquid-liquid condensates or more solid assemblies depending on the interaction affinities. The team also showed that assemblies that had already formed could be reconfigured by mixing in new components.

This work was supported by EMBO long-term fellowships, Human Frontiers Science Program long-term fellowships, a Washington Research Foundation Innovation fellowship, and by NIH, DARPA, HHMI, Open Philanthropy Project, The Audacious Project, and Eric and Wendy Schmidt by recommendation of the Schmidt Futures. All data are available in the main text or supplementary materials. Design scripts, protein sequences, design models, and models of assemblies are also available through Zenodo.

Deep learning dreams up new protein structures

Just as convincing images of cats can be created using artificial intelligence, new proteins can now be made using similar tools. In a new report in Nature, we describe the development of a neural network that “hallucinates” proteins with new, stable structures.

“For this project, we made up completely random protein sequences and introduced mutations into them until our neural network predicted that they would fold into stable structures,” said co-lead author Ivan Anishchenko, PhD, an acting instructor in the Baker lab at the Institute for Protein Design. “At no point did we guide the software toward a particular outcome — these new proteins are just what a computer dreams up.”

Lead authors Ivan Anishchanka, PhD, and Sam Pellock, PhD.

In the future, it should be possible to steer the artificial intelligence so that it generates new proteins with useful features. “We’d like to use deep learning to design proteins with function, including protein-based drugs, enzymes, you name it,” said co-lead author Sam Pellock, a postdoctoral scholar in the Baker lab.

The research team, which included scientists from UW Medicine, Harvard University, and Rensselaer Polytechnic Institute (RPI), generated two thousand new protein sequences that were predicted to fold. Over 100 of these were produced in the laboratory and studied. Detailed analysis on three such proteins confirmed that the shapes predicted by the computer were indeed realized in the lab.

“Our NMR studies, along with X-ray crystal structures determined by the University of Washington team, demonstrate the remarkable accuracy of protein designs created by the hallucination approach”, said co-author Theresa Ramelot, a senior research scientist at RPI in Troy, New York.

Gaetano Montelione, a co-author and professor of chemistry and chemical biology at RPI, notes “The hallucination approach builds on observations we made together with the Baker lab revealing that protein structure prediction with deep learning can be quite accurate even for a single protein sequence with no natural relatives. The potential to hallucinate brand new proteins that bind particular biomolecules or form desired enzymatic active sites is very exciting”.

Hallucination model of design 0738 (left) and the crystal structure of the surface-modified 0738_mod (right)

“This approach greatly simplifies protein design,” said senior author David Baker. “Before, to create a new protein with a particular shape, people first carefully studied related structures in nature to come up with a set of rules that were then applied in the design process. New sets of rules were needed for each new type of fold. Here, by using a deep-learning network that already captures general principles of protein structure, we eliminate the need for fold-specific rules and open up the possibility of focusing on just the functional parts of a protein directly.”

“Exploring how to best use this strategy for specific applications is now an active area of research, and this is where I expect the next breakthroughs,” said Baker.

Funding was provided by the National Science Foundation (1937533, MCB2032259), National Institutes of Health (DP5OD026389, GM120574, P30GM124165, S10OD021527), Department of Energy (DE-AC02-06CH11357) Open Philanthropy, Eric and Wendy Schmidt by recommendation of the Schmidt Futures program, Audacious Project, Washington Research Foundation, Novo Nordisk Foundation, and Howard Hughes Medical Institute. The authors also acknowledge computing resources from the University of Washington and Rosetta@Home volunteers.

Deep learning reveals how proteins interact

A team led by scientsts in the Baker lab has combined recent advances in evolutionary analysis and deep learning to build three-dimensional models of how most proteins in eukaryotes interact. This breakthrough has significant implications for understanding the biochemical processes that are common to all animals, plants, and fungi. This open-access work appears in Science.

“To really understand the cellular conditions that give rise to health and disease, it’s essential to know how different proteins in a cell work together. In this paper, we provide detailed information on protein interactions for nearly every core process in eukaryotic cells. This includes over a hundred interactions that have never been seen before.”

Co-lead author Ian Humphreys, a graduate student in the Baker lab.

Proteins are the workhorses of all cells, but they rarely act alone. Different proteins often must fit together to form precise “complexes” that carry out specific tasks, including reading genes, digesting nutrients, and responding to signals from neighboring cells and the outside world. When protein complexes malfunction, disease can result.

“This work shows that deep learning can now generate real insights into decades-old questions in biology —not just what a particular protein looks like, but also which proteins come together to interact,” said senior author Qian Cong, an assistant professor in the department of biophysics at the University of Texas Southwestern Medical Center.

To exhaustively map the interactions that give rise to protein complexes, a team of structural biologists from UW Medicine, University of Texas Southwestern Medical Center, Harvard University, and other several institutions examined all known gene sequences in yeast. Using advanced statistical analyses, they identified pairs of genes that naturally acquire mutations in a linked fashion. They reasoned that such shared mutations are a sign that the proteins encoded by the genes must physically interact.

The researchers also used new deep-learning software to model the three-dimensional shapes of these interacting proteins. RoseTTAFold, invented at UW Medicine, and AlphaFold, invented by the Alphabet subsidiary DeepMind, were both used to generate hundreds of detailed pictures of protein complexes.

“As computer methods become more powerful, it is easier than ever to generate large amounts of scientific data, but to make sense of it still requires scientific experts,” said senior author David Baker, director of the Institute for Protein Design. “So we recruited a village of expert biologists to interpret our 3D protein models. This is community science at its best.”

The hundreds of newly identified protein complexes provide rich insights into how cells function. For example, one complex contains the protein RAD51, which is known to play a key role in DNA repair and cancer progression in humans. Another includes the poorly understood enzyme glycosylphosphatidylinositol transamidase, which has been implicated in neurodevelopmental disorders and cancer in humans. Understanding how these and other proteins interact may open the door to the development of new medications for a wide range of health disorders.

The protein structures generated in this work are available to download from the ModelArchive. The authors thank and remember John Westbrook at the Protein Data Bank for his support in establishing formats and software code to allow efficient deposition of the models into the archive; John sadly passed during the preparation of this manuscript.

The project was led by Ian Humphreys, Aditya Krishnakumar, and Minkyung Baek at UW Medicine as well as Jimin Pei at the University of Texas Southwestern Medical Center. Collaborating institutions include UW Medicine, UT Southwestern, Harvard University, Wayne State University, Cornell University, MRC Laboratory of Molecular Biology, Memorial Sloan Kettering Cancer Center, Gerstner Sloan Kettering Graduate School of Biomedical Sciences, Fred Hutchinson Cancer Research Center, Columbia University, University of Würzburg, St Jude Children’s Research Hospital, FIRC Institute of Molecular Oncology, and Istituto di Genetica Molecolare, Consiglio Nazionale delle Ricerche.

This work was supported by Microsoft, Amgen, Southwestern Medical Foundation, The Washington Research Foundation, Howard Hughes Medical Institute, National Science Foundation (DBI 1937533), National Institutes of Health (R35GM118026, R01CA221858, R35GM136258, R21AI156595), UK Medical Research Council (MRC_UP_1201/10), HHMI Gilliam Fellowship, the Deutsche Forschungsgemeinschaft (KI-562/11-1, KI-562/7-1), AIRC investigator and the European Research Council Consolidator (IG23710 and 682190), Defense Threat Reduction Agency (HDTRA1-21-1-0007), Cancer Prevention and Research Institute of Texas (RP210041), and National Energy Research Scientific Computing Center.

RoseTTAFold: Accurate protein structure prediction accessible to all

Today we report the development and initial applications of RoseTTAFold, a software tool that uses deep learning to quickly and accurately predict protein structures based on limited information. Without the aid of such software, it can take years of laboratory work to determine the structure of just one protein. With RoseTTAFold, a protein structure can be computed in as little as ten minutes on a single gaming computer. This work was led by Baker lab postdoctoral scholar Minkyung Baek, Ph.D.

RoseTTAFold is a “three-track” neural network, meaning it simultaneously considers patterns in protein sequences, how a protein’s amino acids interact with one another, and a protein’s possible three-dimensional structure. In this architecture, one-, two-, and three-dimensional information flows back and forth, allowing the network to collectively reason about the relationship between a protein’s chemical parts and its folded structure.

As reported in Science, our team used RoseTTAFold to compute hundreds of new protein structures, including many poorly understood proteins from the human genome. We also generated structures directly relevant to human health, including for proteins associated with problematic lipid metabolism, inflammation disorders, and cancer cell growth. And we show that RoseTTAFold can be used to build models of complex biological assemblies in a fraction of the time previously required.

“In just the last month, over 4,500 proteins have been submitted to our new web server, and we have made the RoseTTAFold code available through the GitHub website. We hope this new tool will continue to benefit the entire research community”.

Lead author Minkyung Baek, Ph.D.

This work was supported in part by Microsoft, Open Philanthropy Project, Schmidt Futures, Washington Research Foundation, National Science Foundation, Wellcome Trust, and the National Institute of Health. A full list of supporters is available in the manuscript.

Nanoparticle flu shot blocks seasonal and pandemic strains

IPD researchers together with collaborators at the National Institutes of Health have developed experimental flu shots that protect animals from a wide variety of seasonal and pandemic influenza strains. The lead vaccine candidate has entered human clinical testing at the NIH. If it proves safe and effective, these next-generation influenza vaccines may replace current seasonal options by providing protection against many more strains that current vaccines do not adequately cover.

A study detailing how the new flu vaccines were designed and how they protect mice, ferrets, and nonhuman primates appears in the March 24 edition of Nature. This work was led by researchers at the Institute for Protein Design and the Vaccine Research Center, which is part of the National Institute of Allergy and Infectious Diseases at the NIH.

Influenza virus causes an estimated 290,000–650,000 deaths per year. Available flu vaccines, which need to be taken seasonally, often fail to protect against many circulating flu strains that cause illness, and the threat of another influenza pandemic looms.

“Most flu shots available today are quadrivalent, meaning they are made from four different flu strains. Each year, the World Health Organization makes a bet on which four strains will be most prevalent, but those predictions can be more or less accurate. This is why we often end up with ‘mismatched’ flu shots that are still helpful but only partially effective,” said lead author Daniel Ellis, a research scientist in the laboratory of Neil King at the IPD.

To create improved influenza vaccines, the team attached hemagglutinin proteins from four different influenza viruses to custom-made protein nanoparticles. This approach enabled an unprecedented level of control over the molecular configuration of the resulting vaccine and yielded an improved immune response compared to conventional flu shots. The new nanoparticle vaccines, which contain the same four hemagglutinin proteins of commercially available quadrivalent influenza vaccines, elicited neutralizing antibody responses to vaccine-matched strains that were equivalent or superior to the commercial vaccines in mice, ferrets, and nonhuman primates. The nanoparticle vaccines—but not the commercial vaccines —also induced protective antibody responses to viruses not contained in the vaccine formulation. These include avian influenza viruses H5N1 and H7N9, which are considered pandemic threats.

“The responses that our vaccine gives against strain-matched viruses are really strong, and the additional coverage we saw against mismatched strains could lower the risk of a bad flu season,” said Ellis.

This study was supported by the National Institutes of Health (R01GM120553, HHSN272201700059C as well as intramural funding to the Vaccine Research Center), Open Philanthropy Project, Audacious Project, Burroughs Wellcome Fund, University of Washington Arnold and Mabel Beckman cryo-EM center, and a Pew Biomedical Scholars Award.

Companion proteins enhance antibody potency

This week we report the design of new proteins that cluster antibodies into dense particles, rendering them more effective. In laboratory testing, such clustered antibodies neutralize COVID-19 pseudovirus, enhance cell signaling, and promote the growth of T cells more effectively than do free antibodies. This new method for enhancing antibody potency may eventually be used to improve antibody-based treatments for a wide range of health disorders.

Antibodies are essential tools in modern medicine, accounting for more than half of all best-selling drugs in recent years. They are used to treat arthritis, cancer, autoimmune disorders, COVID-19, and much more. In 2019, the market for antibody-based technologies reached $150 billion.

Antibodies are typically free-floating proteins. They function by binding to a specific molecular target, which then becomes either activated or inactivated. Virus-neutralizing antibodies that adhere to the surface of the coronavirus can render the pathogen inert. Signaling antibodies that latch onto human cell receptors can alter cellular communication, metabolism, and even gene expression in profound ways.

“We knew that clustering other kinds of signaling proteins can greatly enhance their effects, but there have not been good ways of clustering antibodies,” said lead-author Robby Divine, a graduate student in the department of biochemistry at UW Medicine. Divine led a team of scientists that used molecular design software to create proteins that recognize and bind to specific surfaces that are common to all human antibodies — the so-called fragment crystallizable, or Fc, region.

“Initially, we were just curious to see if we could build proteins that would grab hold of antibodies,” said Divine. With continued bioengineering, the team eventually created proteins that not only bound to antibodies but also assembled them into dense, spherical nanoparticle structures. They call these structures ‘antibody nanocages.’

“Some of the first cages we made would only grab two antibodies per cluster, but we later created cages that could bind six, 12, or even 30. And we quickly found that any antibody we tested could pretty easily be put into a nanocage. It was surprising to see how generic this approach was.”

Lead author Robby Divine, PhD

So far, antibody nanocages have yielded promising results in various laboratory tests. Certain antibodies known to neutralize the COVID-19 coronavirus become seven-fold more potent when formulated into a nanocage. Other antibodies that induce signaling of a protein called CD40 in mammalian cells functioned around 20-fold more potently when formulated into a nanocage, allowing for 20-fold lower dosing to achieve identical signaling results. And when the once-promising anti-cancer antibody conatumumab was formulated into a nanocage, it was able to potently trigger cell death in lab-grown cancer cells. The same antibody alone did not promote cancer cell death, even at the highest concentrations tested. More experiments will be needed to establish whether these trends hold in animal testing.

“The most exciting aspect of this technology is that it is so simple to swap different antibodies into the assemblies. We envision many new treatments emerging from this one common tool,” said senior author David Baker, professor of biochemistry and director of the UW Medicine Institute for Protein Design.

This work was led by UW Medicine and included researchers from the Benaroya Research Institute and Fred Hutchinson Cancer Research Center, Seattle, WA, USA; and Tehran University of Medical Sciences, Tehran, Iran. This work was supported by the National Institutes of Health, National Science Foundation, Howard Hughes Medical Institute, Washington Research Foundation, Audacious Project, Nordstrom-Barrier Directors Fund, Washington State General Operating Fund, Wu Tsai Translational Investigator Fund, Nan Fung Life Sciences Translational Investigator Fund, Fred Hutch COVID-19 Research Fund, a Pew Biomedical Scholars Award, and a Burroughs Wellcome Investigators award.

A deep learning approach to protein design

Scientists at the IPD and Ovchinnikov lab at Harvard have applied deep learning to the challenge of protein design, yielding a new way to quickly create protein sequences that fold up as desired. This breakthrough has broad implications for the development of protein-based medicines and vaccines.

Computational protein design has primarily focused on finding amino acid sequences that encode very low energy target structures. However, what is most relevant during folding is not the absolute energy of the folded state, but the energy difference between the folded state and the lowest-lying alternative states. In a new publication in PNAS, a team led by IPD postdoctoral scholars Christoffer Norn and Basile Wicky describe a deep learning approach that captures the entire folding landscape. They also show that it can enhance current protein design methods.

Read the paper: Norn C, Wicky BIM, et al. Protein sequence design by explicit energy landscape optimization. PNAS. March 2021  [PDF

This work builds on a recently described convolutional neural network called trRosetta that predicts residue-residue distances and orientations from input sets of aligned sequences. Combining these predictions with Rosetta energy minimization yielded excellent predictions of structures in benchmark cases. While co-evolution between pairs of positions in the input multiple sequence alignments is critical for accurate prediction of structures of naturally occurring proteins, it is not necessary for more “ideal” designed proteins: accurate predictions of the latter can be obtained from single sequences.

Because distance and orientation predictions are probabilistic, the team rationalized that they might inherently contain information about alternative conformations, and thus provide more information about design success than classical energy calculations. Moreover, because these predictions can be obtained rapidly for an input sequence on a single GPU, they reasoned that it should be possible to use the network to directly design sequences that fold into a desired structure by maximizing the probability of the observed residue-residue distances and orientations versus all others.

In stark contrast to energy-based sequence design approaches which have characterized the field to date, sequence design using trRosetta has the remarkable ability to capture properties of the entire energy landscape and consider alternative states that can reduce the occupancy of the desired target structure. Such implicit considerations of the full landscape are almost impossible to achieve with atomistic models without employing extremely CPU-intensive calculations, including large-scale Rosetta ab initio structure prediction and molecular dynamics simulations on very long time-scales. On the other hand, because of the lower resolution of the trRosetta method, it is less accurate in the immediate vicinity of the folded structure. Integration of trRosetta design into Rosetta all-atom calculations combines the strong features of both approaches. More generally, this work demonstrates how deep-learning methods can complement detailed physically based models by capturing higher-level properties normally only accessible through large-scale simulations.

This project was supported by the National Institutes of Health, Howard Hughes Medical Institute, Audacious Project, and by Eric and Wendy Schmidt, by recommendation of the Schmidt Futures program. Christoffer Norn is supported by a grant from the Novo Nordisk Foundation. Basile Wicky is an EMBO long-term fellow. The source code for the study is available at

Design of transmembrane beta barrels

Biochemists have created barrel-shaped proteins that embed into lipid membranes, expanding the bioengineering toolkit.

In a milestone for biomolecular design, a team of scientists has succeeded in creating new proteins that adopt one of the most complex folds known to molecular biology. These designer proteins were shown in the lab to spontaneously fold into their intended structures and embed into lipid membranes. Appearing in the journal Science [PDF], this research opens the door to the construction of custom nanoscale tools for advanced filtration and DNA sequencing.

“Right now scientists all over the world are using protein nanopores as part of their effort to sequence genetic material from the pandemic coronavirus and discover mutant strains. For this project, we wanted to design new nanopore proteins completely from scratch that could serve as starting points for a wide range of future applications, including improved DNA sequencing” said lead author Anastassia Vorobieva, a recent postdoctoral scholar in the laboratory of David Baker, director of the Institute for Protein Design at the University of Washington School of Medicine.

Bacteria are encased in a very specialized membrane, called the outer membrane, which protects them from the outside world. Proteins that embed into these membranes facilitate the movement of specific chemicals into and out of the cell. Such natural protein pores share a similar nanoscale structure: a flat sheet of protein that curls in on itself to form a barrel, through which other molecules — including nutrients, vitamins, and even strands of DNA — can pass. This is known as a transmembrane beta-barrel.

To create new transmembrane beta-barrels, Vorobieva and colleagues used molecular design software to draft possible structures. Though they drew inspiration from proteins found throughout the living world, they arrived at sequences that differ from any known before. Their most successful designer proteins contain eight ribbon-like strands that fold into a compact barrel structure that stands just three nanometers tall.

“We began with a relatively simple notion about what would make the proteins fold. But when we tested these initial hypotheses, nothing worked at all. That was very frustrating. We didn’t assume we would get it right the first time, but we did think we could get some information back that would tell us how to move forward. Instead, I had to go back and look carefully at how nature solves this problem. The key was to try to detect patterns in those proteins. It was a really difficult thing to do.”

First author Anastassia Vorobieva, PhD, a recent postdoctoral scholar at the IPD

Researchers in the laboratory of Sheena Radford, Astbury professor of biophysics at the Astbury Centre for Structural Molecular Biology at the University of Leeds, tested whether improved versions of the designer proteins could embed into artificial lipid membranes. They found that they could do so efficiently without the help of any accessory proteins. This is in marked contrast to how natural transmembrane beta barrels fold.

“These designed proteins are interesting from a basic science perspective because they have no evolutionary history,” said Radford, a specialist in protein folding. “By studying them, we can discover some of the essential features that enable transmembrane beta-barrel proteins to fold into a membrane.”

Binyong Liang, an assistant professor working within the laboratory of Lukas Tamm at the University of Virginia School of Medicine, used nuclear magnetic resonance to confirm that the new barrels folded as intended.

This work is the latest achievement in the rapidly progressing field of protein design. In recent years, scientists at the Institute for Protein Design have created innovative vaccines, experimental cancer treatments, and sensors capable of detecting antibodies against COVID-19. The ability to design new proteins from scratch with new functions has implications for diagnosing and treating a wide range of diseases, as well as for materials science.

“With this type of research, it helps to understand a bit about how evolution works at the molecular level, but we are also trying to see beyond that. That’s really the challenge of protein design,” said lead author Vorobieva.

The paper, “De novo design of transmembrane beta-barrels,” appears in the February 19 edition of Science. The research team included scientists from UW Medicine, University of Virginia School of Medicine, University of Leeds, Johns Hopkins University, and The Ohio State University. This work was supported by the National Institutes of Health, Howard Hughes Medical Institute, Fulbright Belgium and Luxembourg, Wellcome Trust, Biotechnology and Biological Sciences Research Council, Medical Research Council, Open Philanthropy Project, Air Force Office of Scientific Research, Nordstrom Barrier Fund, and Eric and Wendy Schmidt by recommendation of the Schmidt Futures program​.

New sensors detect coronavirus proteins and antibodies

This week we report [PDF] a new way to detect the virus that causes COVID-19, as well as antibodies against it. Scientists at the Institute for Protein Design have created protein-based sensors that glow when mixed with components of the virus or specific antibodies. This breakthrough could enable faster and more widespread testing in the near future.

To diagnose coronavirus infection today, most medical laboratories rely on a technique called RT-PCR, which amplifies genetic material from the virus so that it can be seen. This technique requires specialized staff and equipment. It also consumes lab supplies that are now in high demand all over the world. Supply-chain shortfalls have slowed COVID-19 test results in the United States and beyond.

To directly detect key proteins that make up the coronavirus without the need for genetic amplification, a team led by IPD bioengineering graduate student Alfredo Quijano-Rubio and IPD postdoctoral scholar Hsien-Wei Yeh used Rosetta to design new LOCKR-based biosensors. These protein-based devices can recognize either a target protein from the virus or antibodies, bind to them, then emit light through a biochemical luciferase reaction.

Artist’s depiction of a LOCKR-based SARS-CoV-2 biosensor.

Antibody testing can reveal whether someone has had COVID-19 in the past. It is being used to track the spread of the pandemic, but it too requires complex laboratory supplies and equipment.

The same team of UW researchers also created biosensors that glow when mixed with COVID-19 antibodies. They showed that these sensors do not react to other antibodies that might also be in the blood, including those that target other viruses. This sensitivity is important for avoiding false positives.

“We have shown in the lab that these new sensors can readily detect virus proteins or antibodies in simulated nasal fluid or donated serum. Our next goal is to ensure they can be used reliably in a diagnostic setting. This work illustrates the power of de novo protein design to create molecular devices from scratch with new and useful functions” said David Baker, professor of biochemistry and director of the Institute for Protein Design.

Beyond COVID-19, the team also showed that similar biosensors could be designed to detect medically relevant human proteins such as Her2 and Bcl-2, as well as a bacterial toxin and antibodies against Hepatitis B virus.

This research was supported by the National Institutes of Health, Howard Hughes Medical Institute, Air Force Office of Scientific Research, The Audacious Project, Eric and Wendy Schmidt by recommendation of the Schmidt Futures, Washington Research Foundation, Nordstrom Barrier Fund, The Open Philanthropy Project, LG Yonam Foundation, BK21 PLUS project of Korea, United World Antiviral Research Network (UWARN) one of the Centers Researching Emerging Infectious Diseases, as well as gift support from Gree Real Estate and “la Caixa” Foundation.

Protein patches boost cell signaling

Today we report the design of a new class of protein material that interacts with living cells without being absorbed by them. These large, flat arrays built from multiple protein parts can influence cell signaling by clustering and anchoring cell surface receptors. This breakthrough could have far-reaching implications for stem cell research and enable the development of new materials designed to modulate the behavior of living systems.

The research, which appears in the January 6 edition of Nature [PDF], was led by the Baker lab at UW and the Derivery lab at the Medical Research Council Laboratory of Molecular Biology in Cambridge, UK. 

Preformed arrays cluster transmembrane proteins in stable assemblies.

Cells commonly terminate signaling by absorbing both an activated receptor and the molecule that stimulated it, targeting both for destruction inside the cell. “This tendency of cells to internalize receptors likely lowers the efficiency of immunotherapies,” said Emmanuel Derivery, assistant professor at the MRC Laboratory of Molecular Biology. “Indeed, when antibody drugs bind their target receptors and then become internalized and degraded, more antibody must always be injected.”

To create a way around this, Baker lab postdoctoral scholar Ariel Ben-Sasson designed new proteins that can assemble into large, flat patches upon an external cue. This molecular scaffolding was then further engineered to display signaling molecules. Graduate student Joseph Watson of the Derivery lab showed that such protein materials could latch onto cells, activate surface receptors, and resist being absorbed by the cell for hours or even days. By swapping out which cell surface receptors were targeted by the patch, the researchers showed that different cell types could be activated.

“We now have a tool that can interact with any type of cells in a very specific way. This is what is exciting about protein engineering — it opens fields that people may not expect.”

Lead author Ariel Ben-Sasson, PhD, postdoctoral scholar at the IPD

“This work paves the way towards a synthetic cell biology, where a new generation of multi-protein materials can be designed to control the complex behavior of cells,” said David Baker, professor of biochemistry and director of IPD.

According to co-author Hannele Ruohola-Baker, professor of biochemistry and associate director of the UW Institute for Stem Cell and Regenerative Medicine, versions of these new materials could eventually help physicians alleviate the dangers of sepsis by controlling the inflammatory response to infection and even enable entirely new ways to treat COVID-19, heart disease, and diabetes, and even mitigate the downstream effects of strokes, including Alzheimer’s disease.  “This breakthrough helps pave the way for the use of synthetic cell biology in regenerative medicine,” said Ruohola-Baker.

This research was supported by the UK Medical Research Council, UK Engineering and Physical Sciences Research Council, Wellcome Trust, Human Frontier Science Program, Howard Hughes Medical Institute, US National Institutes of Health, US Department of Energy Office of Basic Energy Sciences Biomolecular Materials Program at Pacific Northwest National Laboratory, Medimmune, and Infinitus.

Design of an ultrapotent COVID-19 vaccine candidate

Today we report in Cell (PDF) the design and initial preclinical testing of an innovative nanoparticle vaccine candidate for the pandemic coronavirus. It produces virus-neutralizing antibodies in mice at levels ten-times greater than is seen in people who have recovered from COVID-19.

Compared to vaccination with the soluble SARS-CoV-2 Spike protein, which is what many leading COVID-19 vaccine candidates are based on, the new nanoparticle vaccine produced ten times more neutralizing antibodies in mice, even at a six-fold lower vaccine dose. The data also show a strong B-cell response after immunization, which can be critical for immune memory and a durable vaccine effect. When administered to a single nonhuman primate, the nanoparticle vaccine produced neutralizing antibodies targeting multiple different sites on the Spike protein. This may ensure protection against mutated strains of the virus, should they arise.

The vaccine candidate was developed using structure-based vaccine design techniques invented at UW Medicine. It is a self-assembling protein nanoparticle that displays 60 copies of the SARS-CoV-2 Spike protein’s receptor-binding domain in a highly immunogenic array. The molecular structure of the vaccine roughly mimics that of a virus, which may account for its enhanced ability to provoke an immune response.

The lead authors of this paper are Alexandra Walls, a research scientist in the laboratory of David Veesler who is an associate professor of biochemistry at the University of Washington School of Medicine; and Brooke Fiala, a research scientist in the laboratory of Neil King who is an assistant professor of biochemistry at the University of Washington School of Medicine and head of vaccine research at the Institute for Protein Design.

“We hope that our nanoparticle platform may help fight this pandemic that is causing so much damage to our world. The potency, stability, and manufacturability of this vaccine candidate differentiate it from many others under investigation.”

Neil King, PhD, head of vaccine design at the IPD and inventor of the computational vaccine design technology used in this work.

Hundreds of candidate vaccines for COVID-19 are in development around the world. Many require large doses, complex manufacturing, and cold-chain shipping and storage. An ultrapotent vaccine that is safe, effective at low doses, simple to produce and stable outside of a freezer could enable vaccination against COVID-19 on a global scale.

“I am delighted that our studies of antibody responses to coronaviruses led to the design of this promising vaccine candidate,” said Veesler, who spearheaded the concept of a multivalent receptor-binding domain-based vaccine.

The lead vaccine candidate is being licensed non-exclusively and royalty-free during the pandemic by the University of Washington. One licensee, ​Icosavax, Inc.,​ a Seattle biotechnology company co-founded in 2019 by King, is currently advancing studies to support regulatory filings and has initiated the U.S. Food and Drug Administration’s Good Manufacturing Practice (GMP). To accelerate progress by Icosavax to the clinic, A​mgen Inc.​, has agreed to manufacture a key intermediate for these initial clinical studies. Another licensee, S​K bioscience Co., Ltd.​, based in South Korea, is also advancing its own studies to support clinical and further development.

This work was supported by the National Institutes of Health, Bill & Melinda Gates Foundation, gifts from Jodi Green and Mike Halperin and from The Audacious Project, as well as other granting agencies.

Antiviral proteins block coronavirus infection in the lab

Today we report in Science [PDF] the design of small proteins that protect cells from SARS-CoV-2, the virus that causes COVID-19. In experiments involving lab-grown human cells, the activity of the lead antiviral candidate produced (LCB1) was found to rival that of the best-known SARS-CoV-2 neutralizing antibodies. LCB1 is currently being evaluated in rodents. 

Coronaviruses are studded with so-called Spike proteins that latch onto human cells, leading to infection. Drugs that interfere with this process may treat or even prevent infection. Researchers at the IPD used computers to design new proteins that bind tightly to SARS-CoV-2 Spike protein, interfering with its ability to infect cells. Beginning in January, over two million candidate Spike-binding proteins were designed on the computer, and over 118,000 were produced and tested in the lab.

“Although extensive clinical testing is still needed, we believe the best of these computer-generated antivirals are quite promising. They appear to block SARS-CoV-2 infection at least as well as monoclonal antibodies but are much easier to produce and far more stable, potentially eliminating the need for refrigeration.”

Longxing Cao, postdoctoral scholar at the IPD

The researchers created antiviral proteins using two approaches. First, a segment of the ACE2 receptor, which SARS-CoV-2 naturally binds to, was incorporated into a series of small protein scaffolds. Second, completely synthetic proteins were designed from scratch. The latter method produced the most potent antivirals, including LCB1, which is roughly six times more potent on a per mass basis than the most effective monoclonal antibodies reported thus far.

This work was conducted by scientists from the University of Washington School of Medicine and Washington University School of Medicine in St. Louis.

“Our success in designing high-affinity antiviral proteins from scratch is further proof that computational protein design can be used to create promising drug candidates,” said senior author and HHMI Investigator David Baker, director of the IPD.

To confirm that the new antiviral proteins attached to the coronavirus Spike protein as intended, the team collected snapshots of the two molecules interacting using cryo-electron microscopy. These experiments were performed by researchers in the laboratories of David Veesler, assistant professor of biochemistry at the University of Washington School of Medicine, and Michael S. Diamond, the Herbert S. Gasser Professor in the Division of Infectious Diseases at Washington University School of Medicine in St. Louis.

This work was supported by the National Institutes of Health, Defense Advanced Research Projects Agency, The Audacious Project at the Institute for Protein Design, Eric and Wendy Schmidt by recommendation of the Schmidt Futures program, Open Philanthropy Project, an Azure computing resource gift for COVID-19 research provided by Microsoft, and the Burroughs Wellcome Fund.

Introducing Co-LOCKR: designed protein logic for cell targeting

In a new paper [PDF] appearing in Science, a team of IPD researchers together with colleagues at UW Medicine and Fred Hutchinson Cancer Research Center demonstrate a new way to precisely target cells — including those that look almost exactly like their neighbors. They designed nanoscale devices made of synthetic proteins that target a therapeutic agent only to cells with specific, predetermined combinations of cell surface markers. 

These ‘molecular computers,’ which are based on the LOCKR system, operate all on their own, searching out the cells that they were programmed to find.

“We were trying to solve a key problem in medicine, which is how to target specific cells in a complex environment. Unfortunately, most cells lack a single surface marker that is unique to just them. So to improve cell targeting, we created a way to direct almost any biological function to any cell by going after combinations of cell surface markers.”

Marc Lajoie, PhD, a lead author of the study and recent Baker lab postdoc

The tool they created is called Co-LOCKR, or Colocalization-dependant Latching Orthogonal Cage/Key pRoteins. It consists of multiple synthetic proteins that, when separated, do nothing. But when the pieces come together on the surface of a targeted cell, they change shape, activating a sort of molecular beacon.

The presence of these beacons on a cell surface can guide a predetermined biological activity — like cell killing — to a specific, targeted cell. 

The team showed that Co-LOCKR can focus the cell-killing activity of CAR T cells. In the lab, they mixed Co-LOCKR proteins, CAR T cells, and a soup of potential target cells — some had just one marker, others had two or three. Only the cells with the predetermined marker combination were killed by the T cells. If a cell also had a predetermined “healthy marker”, that cell would be spared.

T cells are extremely efficient killers, so the fact that we can limit their activity on cells with the wrong combination of antigens yet still rapidly eliminate cells with the correct combination is game-changing.”

Alexander Salter, another lead author of the study and MD-PhD student at Fred Hutch.

This cell-targeting strategy relies entirely on proteins, which sets it apart from most other methods that rely on engineered cells and operate on slower timescales.

This research was conducted at the University of Washington School of Medicine Institute for Protein Design, the Immunotherapy Integrated Research Center at Fred Hutchinson Cancer Research Center, and the University of Washington Department of Bioengineering.

The co-lead authors of this work are Marc J. Lajoie (supported by a Washington Research Foundation Innovation Postdoctoral Fellowship and a Cancer Research Institute Irvington Fellowship from the Cancer Research Institute), Scott E. Boyken (supported by the Burroughs Wellcome Fund Career Award at the Scientific Interface), and Alexander I. Salter (supported by the Hearst Foundation and Fred Hutchinson Cancer Research Center Interdisciplinary Training Grant in Cancer Research). This work was also supported by the National Institutes of Health (R01 CA114536, NIGMS T32GM008268, 1R21CA232430-01, T32CA080416), the National Science Foundation (CHE-1629214), the Defense Threat Reduction Agency (HDTRA1-18-1-0001), the Nordstrom Barrier IPD Directors Fund, the Hearst Foundation, the Washington Research Foundation and Translational Research Fund, the Howard Hughes Medical Institute, the Open Philanthropy Project, and The Audacious Project organized by TED.

De novo nanoparticles as vaccine scaffolds

IPD researchers have developed a new vaccine design strategy that could confer improved immunity against certain viruses, including those that cause AIDS, the flu, and COVID-19. Using this technique, viral antigens are attached to the surface of self-assembling, de novo designed protein nanoparticles. This enables an unprecedented level of control over the molecular configuration of the resulting vaccine. This research, which includes collaborative pre-clinical evaluation of initial vaccines in animals, is detailed in three new papers published on August 4.

The first paper, published in the journal eLife, describes the overall vaccine design strategy and how it was used to create vaccine candidates for three important viruses: HIV, RSV, and influenza.

“One of the things we found in this study was that putting the same viral antigen on different nanoparticles alters which regions antibodies can see. This can be used to bias the immune response towards certain regions of an antigen that confer greater protective immunity.” 

George Ueda, lead author and IPD translational postdoctoral scholar.

The second paper, published in PLOS Pathogens, looks at how one of the new HIV vaccine nanoparticles performed in rabbits. A team led by Aleks Antanasijevic and Andrew Ward at Scripps Research found that repeated immunization of the vaccine resulted in a higher proportion of neutralizing antibodies compared to immunization with the same antigen not displayed on the nanoparticle.

The third paper, published in npj Vaccines, looks at how one of the HIV vaccine nanoparticles circulates through the body of rhesus macaques. A team led by Jacob Martin and Darrell Irvine at MIT found that after three days, it became concentrated in lymph node tissues, which is where B cells learn how to fight infection. This may account in part for the observed enhanced immunity.

“Simply injecting an antigen is not necessarily enough to confer a protective immune response. Our goal was to create new protein-based vaccines that mimic the repetitive and spiky shape of a virus because this can drive a more protective immune response. What we found in this study was that the nanoparticle vaccines are also retained better in lymph nodes than antigen alone.” said Ueda.

Relevance for COVID-19

The team chose to focus on HIV, RSV, and influenza because those viruses all contain surface proteins with similar shapes — trimers. The virus that causes COVID-19 also contains a trimeric surface protein. Efforts are now underway at UW Medicine and at the National Institutes of Health Vaccine Research Center to develop nanoparticle vaccines against COVID-19 using this new strategy.

“We have found that the two-component nanoparticles we’ve been designing can be used to improve the potency of antigens from a number of important pathogens, including SARS-CoV-2. We’re convinced that they are a robust and versatile platform for designing nanoparticle vaccines.”

Neil King, head of vaccine design at the IPD.

This collaborative research was led by UW Medicine, Scripps Research, and the Koch Institute for Integrative Cancer Research at MIT. It also included researchers from Cornell University,  Emory University, University of Amsterdam, University of Southampton, the Frederick National Laboratory for Cancer Research sponsored by the National Cancer Institute, the Lawrence Berkeley Laboratory, and the National Institute of Allergy and Infectious Diseases at the National Institutes of Health.

This work was supported by the Bill and Melinda Gates Foundation and the Collaboration for AIDS Vaccine Discovery; the National Institute of Allergy and Infectious Diseases Center for HIV/AIDS Vaccine Immunology and Immunogen Discovery, Center for HIV/AIDS Vaccine Development; and by the National Science Foundation; and by The Audacious Project; and by the Howard Hughes Medical Institute. This work was also supported by the European Union’s Horizon 2020 research and innovation program. This work was partially funded by IAVI with the generous support of USAID, Ministry of Foreign Affairs of the Netherlands, and the Bill & Melinda Gates Foundation.

De novo design of protein logic gates

The same basic tools that allow computers to function are now being used to control life at the molecular level, with implications for future medicines and synthetic biology. 

Together with collaborators, we have created artificial proteins that function as molecular logic gates. These tools, like their electronic counterparts, can be used to program the behavior of more complex systems. The team showed that the new designer proteins can regulate gene expression inside human T-cells, a development that may improve the safety and durability of future cell-based therapies.

“Bioengineers have made logic gates out of DNA, RNA and modified natural proteins before, but these are far from ideal. Our logic gates built from de novo designed proteins are more modular and versatile, and can be used in a wide range of biomedical applications” said David Baker.

Whether electronic or biological, logic gates sense and respond to signals in predetermined ways. One of the simplest is the AND gate; it only produces output when one input AND another are present. For example, when typing on a keyboard, pressing the Shift key AND the A key produces an uppercase letter A. Logic gates made from biological parts aim to bring this level of control into bioengineered systems. With the right gates operating inside living cells, inputs such as the presence of two different molecules — or one and not the other — can cause a cell to produce a specific output, such as activating or suppressing a gene.

Zibo Chen, PhD, was a recent recipient of the 2019 Science & SciLifeLab Prize for Young Scientists.

“The whole Apollo 11 Guidance Computer was built from electronic NOR gates,” said lead author Zibo Chen, a recent UW graduate student. “We succeeded in making protein-based NOR gates. They are not as complicated as NASA’s guidance computers, but nevertheless are a key step toward programming complex biological circuits from scratch.”

Recruiting a patient’s own immune cells in the fight against cancer has worked for certain forms of the disease. But targeting solid tumors with this so-called CAR-T cell therapy approach has proven challenging. Scientists think part of the reason why has to do with T cell exhaustion. Genetically altered T cells can only fight for so long before they stop working, but with protein logic gates that respond to exhaustion signals, the team from UW Medicine hopes to prolong the activity of CAR T cells.

“Longer-lived T cells that are better programmed for each patient would mean more effective personalized medicine,” said Chen.

This work was led by researchers at the UW Medicine Institute for Protein Design. It also included biochemists from Northwestern University, The Ohio State University, Altius Institute for Biomedical Sciences and UC San Francisco.

Read the full report in Science   PDF

Designing shape-shifting proteins

Today we report the design of protein sequences that adopt more than one well-folded structure, reminiscent of viral fusion proteins. This research moves us closer to creating artificial protein systems with reliable moving parts.

In nature, many proteins change shape in response to their environment. This plasticity is often linked to biological function. While computational protein design has been used to create molecules that fold to a single stable state and to re-engineer natural proteins to alter their dynamics or fold, the design from scratch of closely related sequences that adopt well-defined but divergent structures has remained an outstanding challenge.

Kathy Wei, Ph.D.

To create shape-shifting proteins, a team led by recent Baker lab postdoc Kathy Wei, Ph.D., began by identifying sets of amino acid sequences predicted to fold into very different structures — in this case, pairs of cylindrical helical bundles with different lengths.

“We knew from the beginning that we wanted a sequence to transform between a short state with helical “arms” pointed “down” and a long state with helical “arms” pointed “up”. The plan was to use established protocols to first design different proteins that are in each of the two states and then mutate the sequences of these two starting points toward each other until we found a sequence that could fold into both states,” said Wei.

After rounds of design on the computer and testing in the lab, the team succeeded in creating a single molecule that could be seen in both states.

“One of the main challenges for this project was finding a way to tell if the proteins took on the shape they were designed to be in. High-throughput screening methods tend to rely on an enzymatic property of a protein. Since these designed proteins only differed in their shapes, we had to use crystallography and NMR to check their folding, which is a slow process and not guaranteed to yield results.”

“While we found a really promising protein sequence that we can measure in both of the designed states, it’s surprisingly much less dynamic than we would’ve expected. Next, we want to understand how to make the conformational changes more dynamic and how to trigger them in a controlled manner.“

The team included scientists from the University of Washington, UC Berkeley, UC Santa Cruz, and Stanford. Their work was supported by the NIH, DOE, HHMI and the Chan Zuckerberg Biohub.

Computational design of closely related proteins that adopt two well-defined but structurally divergent folds. Kathy Y. Wei, Danai Moschidi, Matthew J. Bick, Santrupti Nerli, Andrew C. McShan, Lauren P. Carter, Po-Ssu Huang, Daniel A. Fletcher, Nikolaos G. Sgourakis, Scott E. Boyken, and David Baker. PNAS.

Read the full report:   PDF

Introducing LOCKR: a bioactive protein switch

Today we report in Nature the design and initial applications of the first completely artificial protein switch that can work inside living cells to modify—or even commandeer—the cell’s complex internal circuitry.

The switch is dubbed LOCKR, short for Latching, Orthogonal Cage/Key pRotein.

“In the same way that integrated circuits enabled the explosion of the computer chip industry, these versatile and dynamic biological switches could soon unlock precise control over the behavior of living cells and, ultimately, our health,” said Hana El-Samad, the Kuo Family Professor of Biochemistry and Biophysics at UCSF and co-senior author of the reports.

LOCKR is made of multiple parts. One chain, called the Cage, sequesters a bioactive peptide. Binding of a second molecule, called the Key, to the Cage causes a change in conformation, exposing the peptide. By swapping out the identify of the caged peptide and by tuning binding affinities, different LOCKR switches can be created for a wide range of signaling outputs.

LOCKR can be used to modify gene expression, redirect cellular traffic, degrade specific proteins, and tightly interface with natural proteins. Together with our colleagues at UCSF, we also built new biological circuits that behave like autonomous sensors. These circuits detect cues from the cell’s internal or external environment and respond by making changes to the cell, just as a thermostat senses ambient temperature and directs a heating or cooling system to shut itself off when a desired temperature is reached.

The lead authors of the reports are Bobby Langan and Scott Boyken of the IPD and Andrew Ng of the UC Berkeley-UCSF Graduate Program in Bioengineering. Both Bobby and Scott have gone on to research positions at Lyell.

“Right now, every cell is responding to its environment,” said Bobby. “Cells receive stimuli, then have to figure out what to do about it. They use natural systems to tune gene expression or degrade proteins, for example.”

Bobby and colleagues set out to create a new way to interface with these cellular systems. They used Rosetta to create and tune LOCKR, testing their tool first in vitro then in vivo.

“LOCKR opens a whole new realm of possibility for programming cells,” said Ng. “We are now limited more by our imagination and creativity rather than the proteins that nature has evolved.”

Read the full reports:

De novo design of bioactive protein switches

Modular and tunable biological feedback control using a de novo protein switch


Coevolution at the proteome scale

Today we report in Science the identification of hundreds of previously uncharacterized protein–protein interactions in E. coli and the pathogenic bacterium M. tuberculosis. These include both previously unknown protein complexes and previously uncharacterized components of known complexes.

This research was led by postdoctoral fellow Qian Cong and included former Baker lab graduate student Sergey Ovchinnikov, now a John Harvard Distinguished Science Fellow at Harvard.

Augmented by sequences from over 40,000 bacterial genomes, the team assessed coevolution between 5.4 million pairs of E. coliproteins. After finding orthologs and building paired alignments, they used a local statistical model to identify over 21,000 putative protein–protein interactions. Three-dimensional models for proteins in each pair were generated and docked, leading to 804 pairs with the strongest evidence for coevolution.

When compared to predictions inferred from high-throughput experimental screening methods, this new coevolution-based method for identifying protein–protein interactions outperforms in both precision and recall on multiple benchmarks.

814 additional pairs were added to this high-confidence set by incorporating protein pairs reported to interact in experimental studies or on the same operon.

“Coevolution has been useful for understanding how specific proteins interact, but we can now use it as a tool for discovery,” said lead author Qian Cong. “We are going to apply this tool to more pathogens, and the human genome. Our success will depend on how much work other scientists put into annotating which parts of the genome are genes and which parts are something else.”

Read the full report:  PDF

Protein arrays on mineral surfaces

Today we report the design of synthetic protein arrays that assemble on the surface of mica, a common and exceptionally smooth crystalline mineral. This work, which was performed in collaboration with the De Yoreo lab at PNNL, provides a foundation for understanding how protein-crystal interactions can be systematically programmed.

Our goal was to engineer artificial proteins to self-assemble on a crystal surface by creating an exact match between the pattern of amino acids in the protein and the atoms of the crystal.

“Biology has an amazing ability to organize matter from the atomic scale all the way up to blue whales,” said co-first author Harley Pyles, a graduate student at the Institute for Protein Design. “Now, using protein design, we can create brand new biomolecules that assemble from atomic- to millimeter-length scales. In this case, mica, a naturally occurring crystal, is acting like a big Lego baseplate on top of which we are assembling new protein architectures.”

Rosetta was used to engineer new proteins with customized patterns of electrical charge on their surfaces — new Lego blocks perfectly matched to the mica baseplate. Different designs formed different patterns when deposited on the mica surface, including crowded wires and highly organized honeycomb-like arrays.

“Even though we designed specific atomic-level interactions, we get these structures, in part, because the proteins are crowded out by the water and are forced to pack together,” said James De Yoreo. “This was unexpected behavior and demonstrates that we need to better understand the role of water in ordering proteins in molecular-scale systems.”

By redesigning parts of the proteins, the team was able to produce honeycomb lattices in which they could digitally tune the diameters of the honeycomb pores by just a few nanometers.

Designing atomically precise filaments and lattices from scratch could unlock entirely novel materials and new strategies for synthesizing semiconductor and metallic nanoparticle circuits for photovoltaic or energy storage applications. Alternatively, the protein honeycombs could be used as extremely precise filters, according to co-first author Shuai Zhang, a postdoctoral researcher at PNNL. “The pores would be small enough to filter viruses out of drinking water or filter particulates out of air,” he said.

Read the full report:  PDF

Protein design by citizen scientists

Citizen scientists can now use Foldit to successfully design synthetic proteins. The initial results of this unique collaboration appear today in Nature.

Brian Koepnick, a recent PhD graduate in the Baker lab, led a team that worked on Foldit behind the scenes, introducing new features into the game that they believed would help players home in on better folded structures.

Players were provided with a poly-isoleucine backbone in a fully extended conformation (60-100 residues in length). They had seven days to fold the backbone into a compact structure and assign a sequence specifying this new structure.

Foldit players produced many creative folds.

Scores in Foldit are calculated using Rosetta. By competing with one another to reach the highest score, Foldit players arrive at virtual proteins with extremely low energies (a high Foldit score corresponds to low protein energy). But since energy alone is not enough for protein design, the Foldit team made adjustments to the Foldit score function. These included requiring the presence of a hydrophobic core, limiting the placement of glycine and alanine, and other side-chain specific terms.

The team experimentally tested 146 Foldit designs. 56 were found to be stable monomers when expressed in E coli. X-ray crystallography and NMR were used to determine the structure of four Foldit designs, which agreed strongly with their design model.

Every step of the way, the team relied on the work of Foldit players to expose problems with the score function. Foldit players are excellent at exploring new kinds of protein folds. For this reason, Foldit players are incredibly helpful for identifying unanticipated weaknesses in Rosetta, and ultimately can improve our understanding of protein folding.

Now that Foldit players can accurately design high-quality proteins from scratch, we can start to challenge Foldit players with more applied protein design problems. Foldit players can now help to design proteins that can assemble into multi-component structures, or that can bind to biological targets as potent medicines, or that can degrade toxic chemicals.

Because Foldit depends on the cooperation and competition of its player community, our scientific ability grows rapidly with the number of Foldit players. We look forward to expanding the Foldit community and recruiting more creative and curious Foldit players!

Read the full manuscript:   PDF

Designed ligands tune cytokine signaling

Today the Baker lab shares some exciting collaborative results of their efforts to design rigid and tunable receptor dimerizers. The first authors of this report are Kritika Mohan, Stanford, and George Ueda, IPD.

From Science:

Exploring a range of signaling
Cytokines are small proteins that bind to the extracellular domains of transmembrane receptors to activate signaling pathways inside the cell. They often act by dimerizing their receptors, and changes in dimer orientation of the extracellular domains can change the signaling output. Mohan et al. systematically explored this tuning effect by designing a series of dimer ligands for the erythropoietin receptor in which they varied the distance and angle between monomers. The topology affected the strength of activation and differentially affected different pathways, which raises the potential for exploiting such ligands in medicinal chemistry.

Read the full report:  PDF

Tunable pH-dependent assemblies

Natural proteins often shift their shapes in precise ways in order to function. Achieving similar molecular rearrangements by design, however, has been a long-standing challenge. Today, a team of researchers lead by scientists at the IPD report in Science the rational design of synthetic proteins that move in response to their environment in predictable and tunable ways.

The team, which included researchers from UW, HHMI, LBNL, and OSU, set out to create pH-responsive oligomers, or pROs, that self-assemble into designed configurations at neutral pH and cooperatively disassemble at lower pH.

The project was lead by Scott Boyken, a recent postdoctoral fellow in the Baker lab, who used a three-step procedure to create the dynamic proteins: first, parametric design was used to create helical-bundle backbones which were then fitted with histidine-rich hydrogen-bond networks using the HBNet algorithm. Finally, for each pRO, the remainder of the new protein sequence was assigned using Rosetta.

“Designing new proteins with moving parts has been a long-term goal of my postdoctoral work,” said Boyken. “Because we designed these proteins from scratch, we were able to control the exact number and location of the histidines. This let us tune the proteins to fall apart at different levels of acidity.”

Scott Boyken, PhD

Collaborators in the Wysocki Lab at OSU used native mass spectrometry to determine the amount of acid needed to cause disassembly of the proteins. They confirmed the design hypothesis that having more histidines at interfaces between the proteins would cause the assemblies to collapse more cooperatively.

Researchers in the Lee lab were able to show that these pROs can disrupt artificial membranes in a pH-dependent manner, mirroring the behavior of natural membrane fusion proteins which also contain amphipathic helices.

Follow-up experiments with the Lippincott-Schwartz lab showed that these proteins can also disrupt endosomal membranes in mammalian cells, making pROs an attractive tool for engineering the delivery of biologics into the cytoplasm through endosomal escape.

Read the full report here:  PDF

Receptor sub-type binders

This week we report in NSMB a combined computational design and experimental selection approach for creating proteins that bind selectively to closely related receptor subtypes. This project was led by Luke Dang, a former Baker lab graduate student, and Yi Miao, a postdoctoral researcher in Christopher Garcia’s lab at Stanford.


To discriminate between closely related members of a protein family that differ at a limited number of spatially distant positions is a challenge for drug discovery. We describe a combined computational design and experimental selection approach for generating binders targeting functional sites with large, shape complementary interfaces to read out subtle sequence differences for subtype-specific antagonism. Repeat proteins are computationally docked against a functionally relevant region of the target protein surface that varies in the different subtypes, and the interface sequences are optimized for affinity and specificity first computationally and then experimentally. We used this approach to generate a series of human Frizzled (Fz) subtype-selective antagonists with extensive shape complementary interaction surfaces considerably larger than those of repeat proteins selected from random libraries. In vivo administration revealed that Wnt-dependent pericentral liver gene expression involves multiple Fz subtypes, while maintenance of the intestinal crypt stem cell compartment involves only a limited subset.

Read the full report here:    PDF

De novo 2D arrays

This week we report in JACS a general approach for designing self-assembling 2D protein arrays. This project was led by Zibo Chen, a recent Baker lab graduate student, and featured collaborators from the, DiMaio, De Yoreo and Kollman labs at UW.


Modular self-assembly of biomolecules in two dimensions (2D) is straightforward with DNA but has been difficult to realize with proteins, due to the lack of modular specificity similar to Watson-Crick base pairing. Here we describe a general approach to design 2D arrays using ​de novo designed pseudosymmetric protein building blocks. A homodimeric helical bundle was reconnected into a monomeric building block, and the surface was redesigned in Rosetta to enable self-assembly into a 2D array in the C 1 2 layer symmetry group. Two out of ten designed arrays assembled to μm scale under negative stain electron microscopy, and displayed the designed lattice geometry with assembly size up to 100 nm under atomic force microscopy. The design of 2D arrays with pseudosymmetric building blocks is an important step toward the design of programmable protein self-assembly via pseudosymmetric patterning of orthogonal binding interfaces.

Read the full report here: PDF

IPD’s first nanoparticle vaccine

Today we report in Cell our first computer-designed nanoparticle vaccine targeting respiratory syncytial virus (RSV).

Millions of children will visit hospitals this year, sickened by RSV. Infection is usually mild, causing only fevers, runny noses and frightened parents. But, in severe cases, barking coughs and painful wheezing can indicate serious respiratory complications, including bronchiolitis and pneumonia.

RSV is the primary cause of pneumonia in children under one and is therefore the leading cause of infant mortality worldwide after malaria. Although virtually every child on Earth will get RSV before the age of three, an estimated 99 percent of RSV deaths occur in developing countries. Despite substantial effort, there is not yet a safe and effective vaccine.

Today, an international team of scientists co-led by researchers at the IPD report in Cell a first-of-its-kind vaccine candidate for RSV. It elicits broadly neutralizing antibodies against respiratory syncytial virus in mice and monkeys, paving the way for human clinical trials.

Vaccines by design

Researchers in the King lab created the new vaccine candidate by fusing a stabilized version of RSV F, a glycoprotein from the virus responsible for membrane fusion, onto their designed two-component protein nanoparticle platform. This yielded a self-assembling vaccine that can be tuned to display a variable number of antigens by stoichiometrically mixing different versions of the purified parts.

Vaccine researchers Brooke Fiala and Neil King in the lab.

Because the core of this vaccine consists of an computer-generated, ultrastable nanoparticle, this multivalent vaccine candidate — dubbed DS-Cav1–I53-50 — is more stable than the trimeric antigen (DS-Cav1) alone. DS-Cav1–I53-50A exhibited no discernible loss in antibody binding as measured by bio-layer interferometry after being stored for two weeks at 37°C relative to -80°C. This extremely high stability may ultimately translate into a vaccine that does not require refrigeration, greatly reducing the cost and complexity of global vaccine distribution.

DS-Cav1 is a prefusion-stabilized variant of the F glycoprotein trimer developed at the National Institute for Allergy and Infectious Disease Vaccine Research Center at the National Institutes of Health that has elicited significantly higher neutralizing antibody titers than postfusion F in naïve mice and non-human primates, cattle​ and, most recently, healthy adult humans. DS-Cav1 is itself being evaluated in a Phase 1 study by NIH as an RSV vaccine candidate. The new nanoparticle vaccine based on DS-Cav1 was ten times more potent than DS-Cav1 alone, suggesting it may translate into a more effective vaccine with more durable protection.

The first of many

“This is the first of many vaccine candidates we have made using this technology,” said senior author Neil King. By swapping out the proteins along the outside of the nanoparticle, King and his team hope to create new vaccines candidates for HIV, malaria, and even cancer.

The RSV vaccine team was led by researchers at UW and the Institute for Research in Biomedicine in Bellinzona, Switzerland. It also included scientists from the Fred Hutch Cancer Research Center in Seattle, USA; the Karolinska Institute in Stockholm, Sweden; the Vaccine Formulation Institute in Godalming, UK; the European Virus Bioinformatics Center in Jena, Germany; the Vaccine Formulation Laboratory at the University of Lausanne, Switzerland; and the ​Institute of Microbiology at ETH Zürich, Switzerland. The project was funded in part by the Bill and Melinda Gates Foundation.

Read the full report here: (PDF)

Potent anti-cancer proteins with fewer side effects

Today we report in Nature the first de novo designed proteins with anti-cancer activity.

These compact molecules were designed to stimulate the same receptors as IL-2, a powerful immunotherapeutic drug, while avoiding unwanted off-target receptor interactions. We believe this is just the first of many computer-generated cancer drugs with enhanced specificity and potency.

“People have tried for 30 years to alter IL-2 to make it safer and more effective, but because naturally occurring proteins tend not to be very stable, this has proved to be very hard to do,” said a lead author of the paper, Daniel-Adriano Silva, an IPD biochemist. “Neo-2/15 is very small and very stable. Because we designed it from scratch, we understand all its parts, and we can continue to improve it making it even more stable and active.”

“Neo-2/15 has therapeutic properties that are at least as good as or better than naturally occurring IL-2, but it was computationally designed to be much less toxic,” said another lead author, Umut Ulge, an internal medicine physician and IPD biochemist.

Daniel, Umut, Carl Walky and Alfredo Rubio from the IPD have started a company to help bring this exciting drug to market. We wish them the very best in their new venture!

Read the full report here: (PDF)

New designer proteins mimic DNA

To close out the year, Baker Lab scientists published a new report describing the creation of proteins that mimic DNA. We believe this breakthrough will aid the creation of bioactive nanomachines.

DNA is a widely used building material at the nanoscale because it is simple and predictable: A pairs with T and C pairs with G. Because of this, DNA strands can be programmed to click together into precise and increasingly complex structures. But DNA has drawbacks. It is not as bioactive as RNA, and not nearly as active as proteins. Bioactive protein assemblies run cells (kinetochores, polymerases, proteasomes, etc). What if designing them was as easy as clicking together DNA?

Using computational design, we created heterodimeric proteins that form double helices with hydrogen-bond mediated specificity. When a pool of these new protein zippers gets melted and then allowed to refold, only the proper pairings form. They are all-against-all orthogonal. With these new tools in hand, we can now begin constructing large protein-based machines that self-assemble in predictable ways.

This project was led by graduate student Zibo Chen and was done in collaboration with the Wysocki Lab at Ohio State University and the Sgourakis Lab at the UC Santa Cruz. The work used support from the SIBYLS program with SAXS and the ALS resources at LBNL, as well as the Argonne Leadership Computing Facility at ANL.

Read the full here: (PDF)

Rolling out new jellies

The basic parts of proteins — helices, strands and loops — can be combined in countless ways. But certain combinations are trickier than others. This week scientists from the IPD, along with collaborators in Brno and Santa Cruz, published the first-ever example of designed non-local beta-strand interactions.

Beta-sheet proteins carry out critical functions in biology, and hence are attractive scaffolds for computational protein design, but the de novo design of all-beta-sheet proteins from first principles has lagged far behind the design of all-alpha or mixed-alpha-beta domains.

Tamuka M. Chidyausiku, a biochemistry graduate student, was one of the project leaders.

Local beta-strand interactions occur when residues near one other hydrogen bond to form compact sheets. To get similar interactions from stretches of residues that are not close in primary sequence, a protein backbone must fold into a complex interwoven shape. The successful design of non-local beta-strand interactions demonstrates a significant advance in our ability to control both fine features (such as precise hydrogen bonding) and global features (such as complex topology) in proteins and opens the door to the design of a broad range of non-local beta-sheet structures.

By studying loops that connect unpaired beta-strands (beta-arches), the team identified a series of structural relationships between loop geometry, side chain directionality and beta-strand length that arise from hydrogen bonding and packing constraints on regular beta-sheet structures. They used these rules to de novo design jellyroll structures with double-stranded beta-helices formed by eight antiparallel β-strands. NMR of a hyperthermostable design closely matched the computational model, demonstrating accurate control over the beta-sheet structure and loop geometry.

Read the full report here: (PDF)

Fluorescent proteins designed from scratch

In the summer of 1961, Osamu Shimomura drove across the country in a cramped station wagon to scoop jellyfish from the docks of Friday Harbor. He wanted to discover what made them glow.

It took Shimomura and other biochemists more 30 years to find a full answer. By then, recombinant DNA technology allowed researchers to clone and characterize the two proteins responsible: aequorin and GFP. The latter would earn Shimomura his share of the 2008 Nobel Prize.

GFP, a 238-residue beta-barrel with a covalently linked chromophore, transformed how scientists study cells and the molecules in them. As a genetic tag, GFP has illuminated the inner workings of human brain cells, bacteria, fungi and more.

This week, scientists from the IPD report in Nature the design of a completely artificial fluorescent beta-barrel protein.

Comparison of of GFP (left) and the new fluorescent protein (right) a, Surface
mesh and ribbon representations. b, Close-up of chromophore
binding interactions.

Many natural proteins evolved to bind small molecules. Reengineering such proteins is rarely straightforward, limiting how they can be applied. The new findings demonstrate that proteins unlike any found in nature can be rationally-designed to bind to and act on specific small molecules with high precision and affinity.

The lead authors of the paper are Jiayi Dou, Ph.D and Anastassia A. Vorobieva, Ph.D., then both senior fellows in the Baker lab.

Anastassia Vorobieva with her son Alexandre (left) and Jiayi Dou (right).

To make the fluorescent protein the researchers had to achieve another first: Creating beta-barrels from scratch. The fold was ideal because one end of its cylindrical shape could be used to stabilize the protein while the other could be used to create a cavity that would serve as the binding site for the target molecule, DFHBI. In nature, beta-barrels proteins bind a wide range of small molecules.

Rosetta was used to design the scaffold de novo. To create the cavity, the team used a new docking algorithm called the “Rotamer Interaction Field” or RIF, developed by William Sheffler, Ph.D., a senior research scientist in the Baker lab, that rapidly identifies all potential structures of cavities that fulfill the prerequisites for binding specific molecules.

The designed protein absorbs blue and emits cyan light. It is stable up to 75°C.

“It worked in bacterial, yeast and mammalian cells,” said Dou, “and being half the size of green fluorescent protein should be very useful to researchers.”

“Equally important,” Baker added, “it greatly advances our understanding of the determinants of protein folding and binding beyond what we have learned from describing existing protein structures.”

Written by Ian Haydon

Read the article here:
View the PDF here: Fluorescent proteins designed from scratch

De Novo Design of Membrane Proteins

It is now possible to create complex, custom-designed transmembrane proteins from scratch !   Today Baker lab members published in Science  “Accurate computational design of multipass transmembrane proteins

Designed Membrane Protein:  (Left) Side view showing the designed membrane protein inside the membrane.  (Right)  Top view of same.

The Abstract reads as follows:

The computational design of transmembrane proteins with more than one membrane-spanning region remains a major challenge. We report the design of transmembrane monomers, homodimers, trimers, and tetramers with 76 to 215 residue subunits containing two to four membrane-spanning regions and up to 860 total residues that adopt the target oligomerization state in detergent solution. The designed proteins localize to the plasma membrane in bacteria and in mammalian cells, and magnetic tweezer unfolding experiments in the membrane indicate that they are very stable. Crystal structures of the designed dimer and tetramer—a rocket-shaped structure with a wide cytoplasmic base that funnels into eight transmembrane helices—are very close to the design models. Our results pave the way for the design of multispan membrane proteins with new functions.





A New World of Designed Macrocycles

Today marks another major step forward for peptide based drug discovery.  IPD researchers report in Science the computational design of a new world of small cyclic peptides, “Macrocycles”,  increasing the number of the known kinds of these molecules by multiple fold.  The conceptual art image below “Illuminating the energy landscape” shows the power of computational design to explore and illuminate structured peptides across the vast energy landscape.

Small peptides have the benefits of small molecule drugs, like aspirin, and large antibody therapies, like rituximab, with fewer drawbacks.  They are stable like small molecules and potent and selective like antibodies.

Image by Vikram Mulligan. Computational design calculations reveal the peptide macrocycle energy landscape.

Abstract reads as follows.

Mixed-chirality peptide macrocycles such as cyclosporine are among the most potent therapeutics identified to date, but there is currently no way to systematically search the structural space spanned by such compounds. Natural proteins do not provide a useful guide: Peptide macrocycles lack regular secondary structures and hydrophobic cores, and can contain local structures not accessible with L-amino acids. Here, we enumerate the stable structures that can be adopted by macrocyclic peptides composed of L- and D-amino acids by near-exhaustive backbone sampling followed by sequence design and energy landscape calculations. We identify more than 200 designs predicted to fold into single stable structures, many times more than the number of currently available unbound peptide macrocycle structures. Nuclear magnetic resonance structures of 9 of 12 designed 7- to 10-residue macrocycles, and three 11- to 14-residue bicyclic designs, are close to the computational models. Our results provide a nearly complete coverage of the rich space of structures possible for short peptide macrocycles and vastly increase the available starting scaffolds for both rational drug design and library selection methods.

Check out these additional news items from UW Medicine, and Science

Read more and download the paper at the Baker lab web site.

Synthetic Nucleocapsids Have Arrived

Published today in Nature, IPD researchers describe the first synthetic protein assemblies — dubbed synthetic nucleocapsids — that encapsulate their own genome and evolve in complex environments.

Computationally Designed Synthetic Nucleocapsid
Computationally Designed Synthetic Nucleocapsid, Illustration by Institute for Protein Design & Cognition Studio

Synthetic nucleocapsids are built to resemble viral capsids and could be used in future to deliver therapeutics to specific cells and tissues. These icosahedral protein assemblies are based off of previously reported results from the Institute for Protein Design.

The image above visualizes the de novo creation of synthetic nucleocapsids from computationally designed proteins and their evolution to acquire properties that could be useful for drug delivery and other biomedical applications. The narrative was designed as a futuristic hologram projection realized through spiraling DNA composed of binary zeros and ones. The projection and computational imagery evoke futuristic technology and design, while calling out natural evolution through the DNA spiral “time-scale” motif. The heads-up display iconography showing a blood bag, mouse, and RNase A convey the challenges we used to evolve the synthetic nucleocapsids. The single net impression of this image is engaging + enlightening and shows that we are entering the next epoch of synthetic biology in which biological systems can be designed and created from scratch.


The challenges of evolution in a complex biochemical environment, coupling genotype to phenotype and protecting the genetic material, are solved elegantly in biological systems by the encapsulation of nucleic acids. In the simplest examples, viruses use capsids to surround their genomes. Although these naturally occurring systems have been modified to change their tropism and to display proteins or peptides, billions of years of evolution have favoured efficiency at the expense of modularity, making viral capsids difficult to engineer. Synthetic systems composed of non-viral proteins could provide a ‘blank slate’ to evolve desired properties for drug delivery and other biomedical applications, while avoiding the safety risks and engineering challenges associated with viruses. Here we create synthetic nucleocapsids, which are computationally designed icosahedral protein assemblies with positively charged inner surfaces that can package their own full-length mRNA genomes. We explore the ability of these nucleocapsids to evolve virus-like properties by generating diversified populations using Escherichia coli as an expression host. Several generations of evolution resulted in markedly improved genome packaging (more than 133-fold), stability in blood (from less than 3.7% to 71% of packaged RNA protected after 6 hours of treatment), and in vivo circulation time (from less than 5 minutes to approximately 4.5 hours). The resulting synthetic nucleocapsids package one full-length RNA genome for every 11 icosahedral assemblies, similar to the best recombinant adeno-associated virus vectors. Our results show that there are simple evolutionary paths through which protein assemblies can acquire virus-like genome packaging and protection. Considerable effort has been directed at ‘top-down’ modification of viruses to be safe and effective for drug delivery and vaccine applications; the ability to design synthetic nanomaterials computationally and to optimize them through evolution now enables a complementary ‘bottom-up’ approach with considerable advantages in programmability and control.

Read more and obtain a PDF of the paper from the Baker lab web site.  Also read additional news items from Science Daily, GeekWire, UW Newsroom, CEN

Designs on New World of Mini-Protein Therapeutics

Mark your calendars!  September 27, 2017 is the day the doors opened to whole new world of targeted therapeutics.  The Baker lab and numerous talented collaborators published in Nature that it is now possible to conduct “Massively parallel de novo protein design for targeted therapeutics”.  Three factors make this possible: Rosetta molecular modeling algorithms for computational protein design, economical computing power, and inexpensive gene write – read technology. Designer therapeutic mini-proteins have arrived!

Artist impression of designed mini-protein binders targeting Influenza hemagglutinin to effectively bind and neutralize the virus.

The group designed and tested 22,660 mini-proteins of 37–43 residues that target influenza haemagglutinin and botulinum neurotoxin B, along with 6,286 control sequences to probe contributions to folding and binding, and identified 2,618 high-affinity mini-binders. Comparison of the binding and non-binding design sets, which are two orders of magnitude larger than any previously investigated, enabled the evaluation and improvement of the computational model. Biophysical characterization of a subset of the binder designs showed that they are extremely stable and, unlike antibodies, do not lose activity after exposure to high temperatures. The designs elicit little or no immune response and provide potent prophylactic and therapeutic protection against influenza, even after extensive repeated dosing.  This design capability opens the door to a whole new future of genetically encoded, tailor made protein therapeutics.  Its a bright new day.

The news of this breakthrough has been highlighted by GEN, CEN, Science Daily and others.






The Matrix of Protein Design

The Matrix movie (1999) depicts a future in which the reality perceived by most humans is actually a computer simulated reality called “the Matrix”.  Published today in Sciencethe Baker lab and collaborators report on a new kind of Matrix –  a new reality for large scale computational protein design which can achieve massive data driven improvements in our ability to design highly stable, small proteins from scratch.

Illustration by Gabe Rocklin


Following the White Rabbit, Postdoctoral fellow Dr. Gabe Rocklin led a group of scientists to design and test over 15,000 new mini-proteins (which do not exist in nature) to see whether they form stable folded structures. Even major protein design studies in the past few years have generally examined only 50 to 100 designs.  Synthetic DNA technology and high throughput screening permitted the group to conduct large-scale testing of structural stability of multitudes of computationally designed proteins.  In turn, this allows them to perform a “Global analysis of protein folding using massively parallel design, synthesis and testing“.  

Through iterative improvements in the design process, the group arrived at 2,788 stable mini-protein structures, which is at least 50-fold more proteins than have ever been characterized from natural sources for similar sized proteins.  Their small size and stability may be advantageous for treating diseases when the drug needs to avoid the immune system and reach the inside of a cell.

The publication Abstract is a step into the Matrix as Morpheus explains,

Proteins fold into unique native structures stabilized by thousands of weak interactions that collectively overcome the entropic cost of folding. Though these forces are “encoded” in the thousands of known protein structures, “decoding” them is challenging due to the complexity of natural proteins that have evolved for function, not stability. Here we combine computational protein design, next-generation gene synthesis, and a high-throughput protease susceptibility assay to measure folding and stability for over 15,000 de novo designed miniproteins, 1,000 natural proteins, 10,000 point-mutants, and 30,000 negative control sequences, identifying over 2,500 new stable designed proteins in four basic folds. This scale — three orders of magnitude greater than that of previous studies of design or folding—enabled us to systematically examine how sequence determines folding and stability in uncharted protein space. Iteration between design and experiment increased the design success rate from 6% to 47%, produced stable proteins unlike those found in nature for topologies where design was initially unsuccessful, and revealed subtle contributions to stability as designs became increasingly optimized. Our approach achieves the long-standing goal of a tight feedback cycle between computation and experiment, and promises to transform computational protein design into a data-driven science.

The research has been recognized by opinion leaders and media outlets as a major step for computational protein design.  See articles in Science, Science Daily, Chemistry WorldPhys.orgGEN, and C&E News.



Big Data Shapes the Fold for of Hundreds of Protein Families

Researchers in the Baker lab at the Institute for Protein Design, working in collaboration with the Joint Genome Institute, published in Science the solved folds and structures for hundreds of protein families.   This “big data” approach to large scale protein structure determination was made possible by a team effort that analyzed billions of gene sequences read out from soil, ocean, and air samples collected around the globe.

Figure 1. Protein Structure Determination from Metagenomic Sequence.

The research has been recognized by numerous opinion leaders and media outlets as an unprecedented breakthrough for protein structure prediction. See articles in The Atlantic, The Economist , Science, GeekWire, and GEN.

How does it work?

As illustrated in Figure 1, the sequencing of DNA from environmental samples produces billions of new protein amino acid sequences. Computer algorithms are used to align the sequences according to their evolutionary history. This allows the discovery of pairs of amino acids that co-evolve. If a change occurs in one amino acid, then a compensatory change is typically observed in another amino acid in the sequence. Co-evolving pairs of amino acids are almost always in close proximity to each other (green and yellow lines) within in the final 3D structure of the protein structure (white backbone).

Why is it important?

With this approach, the team produced reliable models for 622 protein families, and discovered more than 100 new protein folds. In addition to resolving the folding structure of a protein, as shown in Figure 2 co-evolution data can also provide data on the dynamic nature of protein structure including transient contacts, protein-protein contacts, and contacts with ligands. Over time, as more environmental DNA sequence data becomes available, we expect to greatly increase our understanding of protein structure, assembly, and function. In turn, we expect this information to enable the design of new proteins with functions.

Figure 2. Important Protein Contacts Inferred from Co-evoling Amino Acid Pairs.

Sharing data.

The Institute for Protein Design believes in sharing its insights with the rest of the world and we have made publicly available the database of protein structures resolved by these methods.


Design of novel cavity containing proteins

An example of computationally designed proteins made of curved beta-sheets and helices forming cavities with different sizes and shapes. (Benjamin Basanta)

The latest paper coming out from the IPD was published today on the Science website. It’s titled “Principles for designing proteins with cavities formed by curved β sheets” with first co-authors Enrique Marcos and Benjamin Basanta, a former and current IPD member, respectively. Other IPD members on the paper include Tamuka Chidyausiku, Gustav Oberdorfer, Daniel-Adriano Silva, Jiayi Dou, and David Baker. Dr. Baker wrote a summary about the publication:

Some of the key functions of the proteins in our bodies and in all living things are to catalyze chemical reactions—speed up the rates by many orders of magnitude-and to sense and respond to small molecules in the body and in the environment.  New proteins that catalyze chemical reactions and/or sense and respond to compounds not found in nature would have wide use in medicine and industry.

Another example of computationally designed proteins made of curved beta-sheets and helices forming cavities with different sizes and shapes (Tamuka Chidyausiku)

Computational protein design can in principle be used to generate such new catalysts and receptors, but a major challenge to accomplishing this has been the inability to design proteins with cavities within which the catalysis or small molecule binding can take place.  This paper describes a general approach for designing proteins with cavities with tunable size and shape. The method opens the door to design of new catalysts and binding proteins [by generating proteins with appropriately sized and shaped cavities to hold the small molecule and lining the cavity with amino acid functional groups to carry out catalysis and/or binding].

Read the UW’s HS NewsBeat write-up here.

Unleashing the Power of Synthetic Proteins

Published today in Science Philanthropy Alliance,  David Baker, Director of the Institute for Protein Design describes how the opportunities for computational protein design are endless — with new research frontiers and a huge variety of practical applications to be explored, from medicine to energy to technology.

This is an exciting time as we are undergoing a technological revolution in protein design—rather than simply tweaking proteins that have come through the evolutionary process, we are becoming able to design new proteins from scratch to solve current challenges.

Computationally Designed Barrel Protein, Image by Possu Huang


Hyper-stable Designed Peptides and the Coming of Age for De Novo Protein Design

Small constrained peptides combine the stability of small molecule drugs with the selectivity and potency of antibody-based therapeutics. However, peptide-based therapeutics have largely remained underexplored due to the limited diversity of naturally occurring peptide scaffolds, and a lack of methods to design them rationally.  New computational design and wet lab methods developed at the Institute for Protein Design have now opened the door to rational design of a whole new world of hyper-stable drug-like peptide structures.

In an article published in Nature this week, Baker lab / IPD scientists and their collaborators describe the development of computational methods for de novo design of constrained peptides with exceptional stabilities. They used these computational methods to design 18-47 residue constrained peptides with diverse shapes and sizes. The designed peptides presented in the paper cover three broad categories:

1) genetically encodable disulfide cross-linked peptides,

2) synthetic disulfide cross-linked peptides with non-canonical sequences, and

3) cyclic peptides with non-canonical backbones and sequences.

Experimentally determined structures for these peptides are nearly identical to their design models.

EHEE Designed Peptide, Visual Illustration by Vikram Mulligan. The molecular surface is shown as a transparent blue shell, and the peptide’s backbone structure is pink. The amino acid’s side chains are white (carbon atoms), blue (nitrogen atoms) and red (oxygen atoms). The crisscrossing bonds that give the peptide its constrained, stable shape are in bright white.

By including D-amino acids (mirror images of the L-amino acids), and thus expanding the palette of building blocks, Baker lab scientists designed peptides in a sequence and structure space sampled rarely by Nature. Indeed, the article describes successful design of a cyclic 2-helix peptide of mix chirality that represents a shape beyond natural secondary- and tertiary structure.

These designed peptides also exhibit exceptional stability to thermal and chemical denaturation, and thus could serve as attractive scaffolds for design of novel peptide-based therapeutics. More broadly, development of this new computational toolkit to precisely design constrained peptides opens the door for “on-demand” development of a new generation of peptide-based therapeutics.  See In the Pipeline.

These and other breakthroughs in computational protein design are also covered in a Nature review article by David Baker, Po-Ssu Huang, and Scott E. Boyken entitled “The coming of age of de novo protein design”, part of special supplement on The Protein World.

Illustrations of designed peptides with different configurations of two structures: tightly wound ribbons and flat, arrow-shaped ribbons.
Illustrations of designed peptides with different configurations of two structures: tightly wound ribbons and flat, arrow-shaped ribbons.

See additional news coverage

HS NewsBeat, Hutch News,

Funding Sources

The National Institutes of Health provided partial support for this work through grants P50 AG005136, T32-H600035., GM094597, GM090205, and HHSN272201200025C.  Additional funding was provided by The Three Dreamers.

Designed Protein Containers Push Bioengineering Boundaries

Earlier this month, Baker lab researchers reported the computational design of a hyperstable 60-subunit protein icosahedron in Nature (Hsia et al); icosahedral protein structures are commonly observed in natural biological systems for packaging and transport (e.g. viral capsids). The described design was composed of 60 trimeric protein building blocks that self-assembled in a nanocage.

In new work published today, Baker lab scientists and collaborators have taken this work to an exciting new level by engineering 120-subunit icosahedral nanocages that self-assemble from not one, but two distinct protein components. The new designed proteins are described in the latest issue of Science in a paper entitled “Accurate design of megadalton-scale multi-component icosahedral protein complexes”.

In this paper, former Baker lab graduate student Jacob Bale, Ph.D. and collaborators describe the computational design and experimental characterization of ten two-component protein complexes that self-assemble into nanocages with atomic-level accuracy. These nanocages are the largest designed proteins to date with molecular weights of 1.8-2.8 megadaltons and diameters comparable to small viral capsids. The structures have been confirmed by X-ray crystallography (see figure). The advantage of a multi-component protein complex is the ability to control assembly by mixing individually prepared subunits. The authors show that in vitro mixing of the designed subunits occurs rapidly and enables controlled packaging of negatively charged GFP by introducing positive charges on the interior surfaces of the two copmonents.

The ability to design, with atomic-level precision, these large protein nanostructures that can encapsulate biologically relevant cargo and that can be genetically modified with various functionalities opens up exciting new opportunities for targeted drug delivery and vaccine design. A link to the paper and additional information is below:

Link to PDF can be found here

Featured article and video in Science magazine:

This protein designer aims to revolutionize medicines and materials” – Science

Other related news items:

Virus-inspired contender design may lead to cell cargo ships” – UW Health Sciences NewsBeat

Large Protein Nanocages Could Improve Drug Design and Delivery” – HHMI News

Biggest Little Self-Assembling Protein Nanostructures Created” – DARPA News and Events


Watch a short video about the designed protein nanocages

See specific descriptions on these nanoparticles from Jacob Bale, Neil King, and Yang Hsia

Nature provides many examples of self- and co-assembling protein-based molecular machines, including icosahedral protein cages that serve as scaffolds, enzymes, and compartments for essential biochemical reactions and icosahedral virus capsids, which encapsidate and protect viral genomes and mediate entry into host cells. Inspired by these natural materials, we report the computational design and experimental characterization of co-assembling two-component 120-subunit icosahedral protein nanostructures with molecular weights (1.8-2.8 MDa) and dimensions (24-40 nm diameter) comparable to small viral capsids. Electron microscopy, SAXS, and X-ray crystallography show that ten designs spanning three distinct icosahedral architectures form materials closely matching the design models. In vitro assembly of independently purified components reveals rapid assembly rates comparable to viral capsids and enables controlled packaging of molecular cargo via charge complementarity. The ability to design megadalton-scale materials with atomic-level accuracy and controllable assembly opens the door to a new generation of genetically programmable protein-based molecular machines.

Reprinted with permission from AAAS
Reprinted with permission from AAAS

Flu Binder Paper Debuts in PLOS Pathogens

Today at 11am, the paper titled “A Computationally Designed Hemagglutinin Stem-Binding Protein Provides In Vivo Protection from Influenza Independent of a Host Immune Response” was published to the PLOS Pathogen website. This paper was contributed to by several IPD members, including Aaron Chevalier, Jorgen Nelson, Lance Stewart, Lauren Carter, and David Baker. The research was performed in collaboration with colleagues in Deborah Fuller’s lab at UW’s Department of Microbiology. You can find the paper here. Scroll down for the official press release.

FluBinder Rendered by Vikram Mulligan, PhD
FluBinder Rendered by Vikram Mulligan, PhD

Fighting Flu with Designer Drugs: A New Compound Given Before or After Exposure Fends Off Different Influenza Strains

A study published on February 4th in PLOS Pathogens reports that a new antiviral drug protects mice against a range of influenza virus strains. The compound seems to act superior to Oseltamivir (Tamiflu) and independent of the host immune response.
Influenza viruses under the microscope look a bit like balls covered with spikes. The spikes are actually two different proteins, hemagglutinin (HA) and neuraminidase (NA). Both proteins consist of an inner stem region (which doesn’t differ much between flu strains), and a highly variable outer blob. The individual variants fall into designated groups, and this is how flu strains are categorized (for example as H1N1, or H3N5).
Ongoing mutations that change the HA and NA blobs are the reason why flu vaccines differ from season to season; they are based on researchers’ best guesses of what next year’s prominent strains will look like. And dangerous pandemic strains often have radically new blobs against which existing immunity is limited.
In the search for drugs that act broadly against different influenza strains, researchers had previously shown that antibodies against the HA stem region can prevent influenza infection. Such antibodies are protective, at least in part, because they activate the host immune response which then destroys the whole HA/antibody complex. The approach, then, depends on a fully functional immune system—which is not present in infants, the elderly, or immune-compromised individuals.
Inspired by the earlier work, Deborah Fuller from the University of Washington in Seattle, USA, who is interested in developing influenza drugs and vaccines, teamed up with David Baker, also at the University of Washington, who is an expert in computational protein design. Together with colleagues, they set out to design small molecules that—like the protective antibodies—bind to the HA stem, and to test whether these small molecules can protect against influenza infection. Designed to mimic antibodies, the small molecules bind the virus in a similar manner. However, because they don’t engage the immune system the way antibodies do, and because of questions of stability and potency, it was not clear whether they would be able to prevent infection in animals, or eventually, in humans.
Before testing their molecules in animals, the researchers optimized their favorite small molecule candidate by systematically generating thousands of versions and testing how tightly they bound HA stems from seven different influenza strains. As they predicted, the resulting molecule, called HB36.6, protected cells against influenza virus infection in vitro (i.e., in test tubes).
The researchers next tested HB36.6 in “challenge experiments” in mice. They gave mice a single intranasal dose of the drug and 2 hours, 24 hours, or 48 hours later injected them with a normally lethal dose of influenza virus. This one-time HB36.6 treatment, when given up to 48 hours before the challenge, conveyed complete protection: All of the treated mice survived and had little weight loss, whereas all untreated control mice died after losing a third of their body weight or more. Intranasal HB36.6 was also able to protect mice after they had been exposed to flu virus, when administered either as a single dose within a day after exposure, or when it was given daily for four days starting 24 hours after exposure.
This protection does not depend on an intact host immune response. When the researchers repeated the challenge experiments in two different immune-deficient mouse strains, they found that HB36.6 can protect these mice as well.
Comparing HB36.6 with Oseltamivir, the researchers found that a single dose of HB36.6 provided better protection than 10 doses (twice daily for 5 days) of Oseltamivir. Furthermore, when they gave a low dose of HB36.6 post-infection (which by itself was not able to afford full protection) together with twice-daily doses of Oseltamivir, all the mice survived, indicating a synergistic effect when the two antiviral drugs are combined.
Their results, the researchers conclude, “show that computationally designed proteins have potent anti-viral efficacy in vivo and suggests promise for development of a new class of HA stem-targeted antivirals for both therapeutic and prophylactic protection against seasonal and emerging strains of influenza”.

Big moves in protein structure prediction and design

[envira-gallery id=”3246″]

Custom design with atomic level accuracy enables researchers to craft a whole new world of proteins

Naturally occurring proteins are the nanoscale machines that carry out essentially all of the critical functions in living things.

While it has been known for over 40 years that the sequence of amino acids completely determines the shape of the protein, it has been very challenging to predict from the amino acid sequence of the protein its three-dimensional structure, and conversely, to come up with brand new amino acid sequences which fold up into hitherto unseen structures.

Over the past months, scientists at the Institute for Protein Design at the University of Washington and the Fred Hutch, along with colleagues at other institutions, have reported advances in two long-standing problem areas related to the construction of new proteins from scratch.

“It has been a watershed year for protein structure prediction and design,” said UW Medicine researcher David Baker, a University of Washington professor of biochemistry, Howard Hughes Medical Institute investigator and head of the Institute for Protein Design.

The protein structure problem is about figuring out how a protein’s chemical makeup predetermines its molecular structure, and in turn, its biological role.   UW researchers have developed powerful new algorithms using co-evolution data from DNA sequences to make unprecedented highly accurate blind ‘ab initio’ structure predictions of large proteins (>200 amino acids in length). This has opened the door to accurate prediction of the structures for hundreds of thousands of newly discovered proteins in the ocean, soil, and gut microbiome.

Equally difficult is the second problem, which is designing amino acid sequences that will fold into brand new protein structures. Breakthroughs demonstrate that it is now possible to make brand new amino acid sequences with exacting precision for folds inspired by the natural world; and more importantly to make amino acid sequences from scratch for totally novel unknown folds, far surpassing what is predicted to occur in natural proteins.

The new proteins are designed with the help of volunteers around the world participating in the Rosetta@Home distributed computing project. The designed amino acid sequences are encoded in synthetic genes, the proteins are produced in the laboratory, and their structures determined with X-ray crystallography.   The computer models in almost all cases match the experimentally determined crystal structures with near atomic level accuracy.

Researchers report new protein designs for barrels, sheets, rings, and screws –all with near atomic level accuracy. This builds on previous reports of designed protein cubes and spheres; providing proof that it is possible to make a totally new class of protein materials.

With these advances in both protein structure prediction and molecular design, Institute for Protein Design researchers hope to build a new world of proteins with exact specifications for performing critically needed tasks in medical, environmental and industrial arenas.

Examples of their goals are nanoscale tools that:

boost the immune response against HIV and other recalcitrant viruses

block the flu virus so that it can’t infect cells

deliver drugs to cancer cells with precision and reduced side effects

stop allergens from causing symptoms

neutralize deposits, called amyloids, thought to damage vital tissues in Alzheimer’s disease

mop up medications in the body as an antidote

fulfill other diagnostic, therapeutic, and clean energy needs

Just as the manufacturing industry was revolutionized by creating interchangeable parts designed to precise specifications, custom designed protein modules with the right twist, turns, and connections for their modular assembly is a bold new direction for biotechnology.

Results providing proof of this possible future have been reported in recent months by researchers the UW Institute for Protein Design in collaboration with researchers at the Fred Hutch, Max Planck Institute for Developmental Biology, Janelia Research Campus, and the Institute for Molecular Science in Japan.



Evolution offers clues to shaping proteins: The function of many proteins tends to stay the same across species, even after their amino acid sequences have changed over billions of years of evolution. Locating co-evolved pairs of amino acids helps calculate their proximity when the molecule folds. UW graduate student Sergey Ovchinnikov applied this co-evolution DNA sequence analysis in an E-Life paper published on September 3, 2015 entitled “Large-scale determination of previously unsolved protein structures using evolutionary information” that illuminated for the first time the structures of 58 families of proteins containing hundreds of thousands of additional structurally related family members.

“This achievement was a grand slam home run in the history of protein structure prediction,” said Baker.



Modular construction of proteins with repeating motifs: Proteins composed of repeated modules, similar to interlocking Lego® blocks, are common in the natural world. Two papers published in the December 16 issue of Nature entitled, “Exploring the repeat protein universe through computational protein design,” and “Rational design of alpha-helical tandem repeat proteins with closed architectures,” shows that existing repeat proteins occupy only a small fraction of the available space, and that it is possible to design totally new proteins with precisely specified geometries that go far beyond what nature has achieved. The work was led by postdoctoral fellows TJ Brunette, Fabio Parmeggiani and Po-Ssu Huang in the lab of David Baker at the University of Washington Institute for Protein Design and Lindsey Doyle and Phil Bradley at the Fred Hutchinson Cancer Research Institute in Seattle.



Barrel-fold design: , Baker lab postdoctoral fellow Po-Ssu Huang, together with Birte Höcker at the Max Planck Institute for Developmental Biology (Tübingen, Germany) discovered the critical but elusive design principles for a barrel-shaped fold underpinning many natural enzyme molecules. The custom designed barrels folds were built at the Institute for Protein Design and reported on November 23, 2015 in the Nature Chemical Biology paper, “De novo design of a four-fold symmetric TIM-barrel protein with atomic-level accuracy.” This breakthrough has opened the door for bioengineers to generate totally new enzymes that speed up chemical reactions by positioning smaller molecules in custom barrel compartments.

Self-assembling apparatus: Naturally occurring ordered protein arrays along a flat plane are found in bacteria, the heart, and other muscles. Overcoming protein interaction complexities, researchers at UW Institute for Protein Design and the Janelia Research Campus of the Howard Hughes Medical Institute succeeded in programming proteins to self-assemble into novel symmetric, 2-dimensional sheets of protein lattice patterns. UW graduate student Shane Gonen in the Baker lab together with his brother Tamir Gonen at Janelia described their work in the June 19, 2015 issue of Science, “Design of ordered two-dimensional arrays mediated by non-covalent protein-protein interfaces.” This research has application in the design self-assembling protein nanomaterials, especially those that could serve as efficient sensors or light harvesters.

Precision sculpting: Protein designers are continuously refining the principles for fashioning ideal protein structures. The latest paper in the October 6, 2015 Proceedings of the National Academy of Sciences, “Control over overall shape and size in de novo designed proteins” further explains methods for systematically varying protein architecture inspired by nature. Such finesse is needed in optimizing designed proteins to take on exact shapes to perform specified functions.   This work has been led by Baker lab graduate student Yu-Ru Lin in collaboration with Nobuyasu Koga at the Institute for Molecular Science in Japan.

Funding Sources:

The Institute of Protein Design has been funded by several federal agencies, including National Institutes of Health, U.S. Department of Energy, National Science Foundation, U.S. Defense Threat Reduction Agency, and U.S. Air Force Office of Scientific Research, the Washington Research Foundation, the Life Sciences Discovery Fund, as well as through private support.

The Institute also depends on a cadre of citizen scientists around the world who volunteer their personal and computer time for protein folding prediction studies through Rosetta@home and the multi-player on-line protein folding game Foldit.


A similar story was also published in UW Health Science’s Newsbeat. Read it here.

CASP3-11 Results Published in E-Life

In the early 1990s, researchers in the field of protein structure prediction were challenged by the problem of how to impartially judge the accuracy of prediction algorithms.  This realization led the protein structure prediction the community to start the Critical Assessment of protein Structure Prediction (CASP), a community-wide, worldwide experiment for protein structure prediction taking place every two years since 1994.   In each CASP, an independent scientific advisory board solicits other researchers to submit experimentally verified, but unpublished, 3D protein structures to CASP.   The linear amino acid sequences of these proteins are then provided to structure prediction researchers, who each have an equal and limited amount of time to submit final structure predictions to the CASP advisory board.  The submitted structure predictions are then compared to the experimentally verified structures using the same metrics for all CASP contributors.  Even though the primary goal of CASP is to help advance methods for identifying protein 3D structure given only its linear amino acid sequence, many view the experiment more as a “world championship” in protein structure prediction.

Over a 16 year period (CASP3-11), the Baker lab has consistently achieved top performance in the hardest category of structure prediction; the “Twilight Zone” where the linear amino acid sequence of the protein shares no discernable relation to any publicly available 3D structure. In 2014 this culminated in our highly accurate blind structure predictions of two large proteins each >200 amino acids in length. Our methods involve using DNA sequence information to help us predict the 3D structures of proteins.

We recently published these results in E-life, and the results are getting significant attention.

You can read the general-public summary here:

Or find the whole thing here:

Also, learn about the first author of the paper, Sergey Ovchinnikov, by watching this interview:

August IPD News Roundup



Following the groundbreaking 2014 Nature paper describing the development of a computational method to design multi-component coassembling protein nanoparticles, comes a publication in Protein Science from Baker lab graduate student Jacob Bale and collaborators. Titled “Structure of a designed tetrahedral protein assembly variant engineered to have improved soluble expression“, the paper reports a variant of a previously low yielding tetrahedral designed material for which structure determination was difficult. The new variant described in the paper had a much improved yield after redesign and the structure obtained agreed with the computational model with high atomic-level accuracy. The methods used here to improve soluble protein yield will be generally applicable to improving the yield of many designed protein nanomaterials.



Congratulations to newly minted PhDs and graduates of the Baker lab Dr. Shawn Yu and Dr. Ray Wang! Both defended their dissertations this month. Dr. Yu gave a talk on “Computational design of interleukin-2 mimetics” and Dr. Wang spoke about “Protein structure determination from cryoEM density maps”. We wish them the best of the luck in their next steps!

The annual RosettaCON meeting was held July 29-Aug1 at the beautiful Sleeping Lady Mountain Resort in Leavenworth, WA. Many IPD scientists attended the conference, heard talks from researchers in Rosetta labs across the country, presented posters on their own research, and socialized with the larger Rosetta community.

New Science paper: Designed 2-D protein arrays

A new Science paper is out from IPD faculty Dr. David Baker titled Design of ordered two-dimensional arrays mediated by noncovalent protein-protein interfaces. Read the abstract below and the article at the link:

We describe a general approach to designing two-dimensional (2D) protein arrays mediated by noncovalent protein-protein interfaces. Protein homo-oligomers are placed into one of the seventeen 2D layer groups, the degrees of freedom of the lattice are sampled to identify configurations with shape-complementary interacting surfaces, and the interaction energy is minimized using sequence design calculations. We used the method to design proteins that self-assemble into layer groups P 3 2 1, P 4 2(1) 2, and P 6. Projection maps of micrometer-scale arrays, assembled both in vitro and in vivo, are consistent with the design models and display the target layer group symmetry. Such programmable 2D protein lattices should enable new approaches to structure determination, sensing, and nanomaterial engineering.

Shown is a 2D array with P 6 symmetry. (Left) the P 6 lattice has two degrees of freedom available for sampling. Sixfolds are represented by hexagons. (Middle) Computationally designed 2D array. (Right) Electron microscopy of designed P 6 array.

Designer Proteins to Target Cancer Cells

BINDI Designer Protein What if scientists could design a completely new protein that is precision-tuned to bind and inhibit cancer-causing proteins in the body? Collaborating scientists at the UW Institute for Protein Design (IPD) and Molecular Engineering and Sciences Institute (MolES) have made this idea a reality with the designed protein BINDI. BINDI (BHRF1-INhibiting Design acting Intracellularly) is a completely novel protein, based on a new protein scaffold not found in nature, and designed to bind BHRF1, a protein encoded by the Epstein-Barr virus (EBV) which is responsible for disregulating cell growth towards a cancerous state. Learn more here.

Accurate Design of Co-Assembling Multi-Component Protein Nanomaterials

TwoComponentthumbnailA new paper is out in the June 5 issue of Nature entitled Accurate design of co-assembling multi-component protein nanomaterials. Scientists at the Institute for Protein Design (IPD), in collaboration with researchers at UCLA and HHMI, have built upon their previous work constructing single-component protein nanocages and can now design and build self-assembling protein nanomaterials made up of multiple components with near atomic-level accuracy. Learn more about this innovative work at this link.

Design of Activated Serine-Containing Catalytic Triads with Atomic-Level Accuracy

Catalytic Serine TriadBaker lab members published in Nature Chemical Biology a paper entitled “Design of activated serine-containing catalytic triads with atomic-level accuracy“, describing the computational design of proteins with idealized serine-containing catalytic triads which can capture and neutralize organophosphate probes.  This work has utility in design of scavengers of environmental toxins. Learn more at this link.

A Computationally Designed Metalloprotein Using an Unnatural Amino Acid

Mulligan metal binder figureWhat if scientists could design proteins to capture specific metals from our environment?  The utility for cleaning up metals from waste water, soils, and our bodies could be tremendous.  Dr. Jeremy Mills and collaborators in Dr. Baker’s group at the University of Washington’s Institute for Protein Design (IPD) address this challenge in the first reported use of computational protein design software, Rosetta, to engineer a new metal binding protein (“MB-07”) which incorporates an “unnatural amino acid” (UAA) to achieve very high affinity binding to metal cations.  Learn more at this link.

Computational Design of a pH Sensitive Antibody Binder

Designed pH-dependent Fc binder (blue) exploits protonation of Histidine-433 (orange) in the Fc portion Immunoglobulin G (IgG, light cyan surface)

Purification of antibody IgG from crude serum or culture medium is required for virtually all research, diagnostic, and therapeutic antibody applications.  Researchers at the Institute for Protein Design (IPD) have used computational methods to design a new protein (called “Fc-Binder”) that is programed to bind to the constant portion of IgG (aka “Fc” region) at basic pH (8.0) but to release the IgG at slightly acidic pH (5.5).  Published on-line at PNAS (Dec. 31, 2013), the paper is entitled Computational design of a pH-sensitive IgG binding protein, co-authored by Strauch, E. – M., Fleishman S. J., & Baker D.  Learn more at this link.

Computational Protein Design To Improve Detoxification Rates Of Nerve Agents

Phosphotriesterase EngineeringV-type nerve agents are among the most toxic compounds known, and are chemically related to pesticides widespread in the environment. Using an integrated approach, described in an ACS Chemical Biology paper entitled Engineering V-type nerve agents detoxifying enzymes using computationally focused libraries, Dr. Izhack Cherny, Dr. Per Greisen, and collaborators increased the rate of nerve agent detoxification by the enzyme phosphotriesterase (PTE) by 5000-fold by redesigning the active site.   Learn more at this link.

Computational Design Of A Protein That Binds Polar Surfaces

Enzyme Inhibitor DesignIn a Journal of Molecular Biology publication entitled Computational design of a protein-based enzyme inhibitor, Dr. Erik Procko and collaborators describe the computational design of a protein-based enzyme inhibitor that binds the polar active site of hen egg lysosome (HEL).  Computational design of a protein that binds polar surfaces has not been previously accomplished.  Learn more at this link.

One Small Molecule Binding Protein, One Giant Leap for Protein Design

Illustration rendering of the digoxigenin binding protein was prepared by Vikram Mulligan

Reported on-line  in Nature (Sept. 4, 2013) researchers at the Institute for Protein Design describe the use of Rosetta computer algorithms to design a protein which binds with high affinity and specificity to a small drug molecule, digoxigenin a dangerous but sometimes life saving cardiac glycoside.  Learn more at this link.



IPD Researchers Publish New Protocols for Preparing Protein Scaffold Libraries for Functional Site Design

Pareto-optimal refinementIPD researchers in the Baker group have published new computational protocols for preparing protein scaffold libraries for functional site design.  Their paper entitled “A Pareto-optimal refinement method for protein design scaffolds improves the search for amino acids with the lowest energy subject to a set of constraints specifying function.  Learn more at this link.

Centenary Award and Frederick Gowland Hopkins Memorial Lecture

Centenary Award LectureDr. David Baker, Director of the IPD delivered the Centenary Award and Frederick Gowland Hopkins Memorial Lecture at  at the MRC Laboratory of Molecular Biology, Cambridge, UK, on December, 13, 2012.   Baker’s lecture entitled “Protein folding, structure prediction and design”  can be read at this published link.  

See:  Baker, D. (2014).  Protein folding, structure prediction and design.. Biochemical Society transactions. 42(2), 225-9.

Proteins Made to Order. Researchers at the IPD Design Proteins from Scratch with Predictable Structures

Design of Ideal Folded ProteinsA team from David Baker’s laboratory at the University of Washington in Seattle have described a set of “rules” for the design of proteins from scratch, and have demonstrated the successful design of five new proteins that fold reliably into predicted conformations.  Their work was published Nature.  Learn more at this link.

Computational Design of Self-Assembling Protein Nanomaterials with Atomic Level Accuracy

Self Assembling NanomaterialsIPD researchers in the Baker group have published in Science a paper entitled “Computational design of self-assembling protein nanomaterials with atomic level accuracy.”  They describe a general computational method for designing proteins that self-assemble to a desired symmetric architecture.  Protein building blocks are docked together symmetrically to identify complementary packing arrangements, and low-energy protein-protein interfaces are then designed between the building blocks in order to drive self-assembly.  Read more at this link.