Chapter 4 - Homology Modeling and Ligand-Based Molecule Design

Abstract

The protein structures available in the Protein Data Bank (PDB) have been experiencing an exponential growth in the past two decades. Not only has the size of the PDB grown more than 60 times in the past 20 years, but also the structures once considered difficult to determine, such as the structures of membranes proteins, have been added to the PDB at an increasing speed. Despite tremendous successes in the field of structural biology, a significant fraction of therapeutically important protein targets have undetermined structures for various reasons. Homology modeling and in silico protein engineering are powerful tools to extend the structural coverage of the druggable genome. Some unknown protein structures are highly homologous to experimentally determined structures, such as Yes1 in the first case study. These protein structures can be predicted with high confidence by using either homology modeling or in silico protein engineering. In the cases when multiple crystal structures of the same sequence similarity to the target proteins are available, how to select the most relevant structures as templates for homology modeling becomes critical. The rule of thumb for template selection is to favor the proteins whose sequences align better with the targets in the important regions, such as the adenosine-5'-triphosphate (ATP)-binding pocket in kinases and transmembrane domains in G protein-coupled receptors (GPCRs). One more important factor has to be taken into consideration in selecting templates for kinase or GPCR targets (i.e., the activation state of the templates). In case study 1, Yes1 structures, both active and inactive conformations, were modeled on the basis of highly homologous kinase cSrc. The binding modes of a number of known Yes1 inhibitors were predicted by molecular docking.