Optimization of BCL::Fold for Protein Folding de novo and with Cryo-EM Restraints

Abstract

One of the basic tenets of biology and biochemistry is that form meets function; this axiom is applicable to all levels of life. Structural data on large classes of proteins, such as membrane proteins (MPs), is largely lacking due to the difficulty of crystallizing MPs, which is particularly problematic when considering that MPs make up about 60% of drug targets and are integral to proper cellular signaling, structure, and transport. Cryo-Electron Microscopy (cryo-EM) can image proteins in their native state. The BioChemical Library (BCL) developed in the Meiler lab includes a protein folding algorithm dubbed BCL::Fold. BCL::Fold can easily be coupled with experimental data, including those from cryo-EM. Recent cryo-EM maps, which can attain sub-4 Å resolution, can discern individual side chain densities. I have developed a method of representing side chains and optimizing the sampling and scoring procedures so that BCL::Fold can leverage this extra level of detail. In addition, I developed an algorithm that can cluster models of similar topologies by looking at the interactions between their SSEs. Effectively clustering topologies streamlines the de novo protein structure prediction process and can optimize the algorithm to select only the most native-like models.