Identifying Transient Binding Pockets in Protein Dynamics for Allosteric Drug Design

Abstract

Allosteric modulators that regulate the activity of the orthosteric ligands are emerging as cutting-edge strategies in drug design. Unlike orthosteric ligands, allosteric modulators bind to topographically distinct domains from those utilized by orthosteric ligands. Allosteric modulators offer unique therapeutic advantages such as high selectivity thereby causing reduced side effects. However, allosteric pockets are difficult to find since they are often formed transiently during the protein dynamics and hence could be absent in the crystal structures. This poses a challenge in designing allosteric modulators using structure based drug design methods that rely solely on crystal structures or homology models. Moreover not all transient pockets are suitable for allosteric modulation, since the allosteric pocket must communicate with the orthosteric site for functional modulation. Thus there is a dire need for novel techniques that utilize information from protein dynamics to detect allosteric sites for drug design. We present here a comprehensive method for designing allosteric modulators using protein dynamics trajectories or NMR data. We have developed a method, VoidVol, to identify transient binding cavities during protein dynamics. Next, using mutual information calculated from the dynamics trajectories, we map the allosteric pipelines communicating with the orthosteric site. The transient pockets having strong allosteric communication with the orthosteric site can be used for screening allosteric modulators. These sites can be further tested for druggability using the program FindBindSite, also developed in our laboratory. The resulting druggable sites can then be used for high-throughput screening of small-molecule database. We have validated this approach using several kinases and GPCRs with known allosteric modulators. The above methodology demonstrates how molecular dynamics can be useful for allosteric drug design. Our method is applicable to any water-soluble or membrane protein with an available crystal structure or homology model.