Mapping Protein Surfaces in Computational Drug Design

点击翻译按钮获取中文摘要

ABSTRACT: The greatest limitation for structure-based drug discovery (SBDD) is the need to neglect water and protein flexibility in most modeling. Here, we outline simulation methods that overcome these limitations. This proposal focuses on developing MixMD, our method for mixed-solvent molecular dynamics (MD). MixMD identifies critical binding sub-sites on protein surfaces (hotspots). Proteins are simulated in a box of explicit water with 5% small, organic probe cosolvents. The waters and probes sample the local environments along the protein surface, and sites with high occupancy of probes are identified as hotspots. MixMD has superior performance over other cosolvent MD methods like MacKerell’s SILCS and Barril’s MDmix. Other methods are plagued by many spurious, misleading, “extra” sites that indistinguishable from real binding sites, which greatly hinders prospective applications. Our long-term goal is to improve SBDD by developing methods that more accurately model protein-ligand binding. Our underlying hypotheses are 1) MixMD’s more complete description of the physics of binding yields better hotspot predictions than traditional SBDD methods and 2) both qualitative and quantitative data from MixMD can be used in SBDD. This proposal outlines two areas for developing MixMD and increasing its impact on SBDD. Specific Aim 1 develops methods for calculating the free energies, entropies, and enthalpies of the hotspot probes. Comparisons will be made between occupancy-based, energy-based, and kinetics-based methods for calculating those key binding properties. Specific Aim 2 will address a series of key challenges in SBDD. First, MixMD will be used to identify bridging water molecules in binding sites. Clearly, hotspot locations ascertain displaceable water, but it is just as important to pinpoint required, bridging waters in binding sites. Second, the accessibility of difficult, cryptic sites will be examined. While mapping the sites, we will determine whether pocket opening and probe binding are sequential events where probes “capture” open states or concerted events where probes “induce” open states by pushing against the malleable torsions of the cryptic pocket. Lastly, MixMD data will be used to predict druggabilities of binding sites. The Non-Redundant set of Druggable and Less Druggable binding sites (NRDLD) will be used to derive a druggability index based on number of hotspots, their affinities, their proximities, and their degree of burial in the protein.