November 13, 2013    Nivón, L. G., Bjelic S., King C., & Baker D. (2013).  Automating human intuition for protein design. Proteins 82:858-66.

Greedy Optimization
A comparison of the raw Rosetta design output with positions adjusted by the greedy algorithm indicated as sticks (left, blue) and the greedy protocol output with those same positions indicated (right, orange) for the most improved case

IPD researchers in the Baker group have improved protein computational design algorithms by “Automating human intuition for protein design.”   In the design of new enzymes and binding proteinshuman intuition is often used to modify computationally designed amino acid sequences prior to experimental characterization. The manual sequence changes involve both reversions of amino acid mutations back to the identity present in the parent scaffold and the introduction of residues making additional interactions with the binding partner or backing up first shell interactions.  To guide the automation of human intuition in the manual stages of protein design, the Baker group assembled a benchmark set of 51 proteins that tests the ability of a method to recapitulate mutation decisions made by human protein designers in realistic novel design situations.  They also developed a new local sequence optimization procedure that uses a greedy algorithm and allows multiple sampling methods to be carried out in serial using metrics too computationally expensive for global sequence design.  The new protocol improves on traditional design methods on the human designer benchmark.