Ryan Lee Hayes

Position: Post-Doctoral Research Associate



I am interested in developing computational protein design methods that are complementary to current methods. Current protein design methods such as Rosetta have enjoyed wide success in the last decade, and can routinely design proteins that fold to a desired structure. Design projects involving ligand binding or catalysis have had some success, but initial designs have low affinity or activity, which must be optimized by experimental techniques such as directed evolution, or fail altogether. More accurate free energy methods are needed to complement existing protein design algorithms. Multisite lambda dynamics (MSLD) is an alchemical free energy method based on molecular dynamics simulations. Because it is based on molecular dynamics, it naturally accounts for important effects such as backbone flexibility, long range electrostatics, and discrete solvent effects which are poorly treated by Rosetta. MSLD is more efficient and scalable than alternative alchemical free energy methods such as FEP because only one simulation is needed for each alchemical transformation, and because MSLD can be generalized to explore combinatorially large sequence spaces by making mutations at multiple sites. Because of its potential in protein design, I began working on MSLD in 2016. My first project involved introducing adaptive landscape flattening (ALF) and the use of soft cores to MSLD. These improvements enabled MSLD to explore much larger perturbations of up to a dozen heavy atoms, and yielded more robust and reproducible results. Next I applied MSLD to calculate protein folding free energies in T4 lysozyme, and achieved excellent agreement with experiment with a mean unsigned error or 1.1 kcal/mol, and a Pearson correlation coefficient of 0.9. I also demonstrated the scalability of MSLD by accurately calculating folding free energies in spaces of up to 240 sequences. I intend to continue pushing MSLD towards use in protein design. Current challenges including obtaining better results in mutations that change the charge of the protein and using MSLD to explore much larger sequence spaces.

Funding from the following agencies: