Enrichment of Druggable Conformations from Apo Protein Structures Using Cosolvent-Accelerated Molecular Dynamics

• Main Authors: Andrew Kalenkiewicz, Barry J. Grant, Chao-Yie Yang
• Format: Article
• Language: English
• Published: MDPI AG 2015-04-01
• Series: Biology
• Subjects: Bcl-2 proteins; protein–protein interactions; principal component analysis; hierarchical clustering; binding site hotspot mapping; Sitemap; Bcl-xL inhibitors; computational docking
• Online Access: http://www.mdpi.com/2079-7737/4/2/344

Here we describe the development of an improved workflow for utilizing experimental and simulated protein conformations in the structure-based design of inhibitors for anti-apoptotic Bcl-2 family proteins. Traditional structure-based approaches on similar targets are often constrained by the sparsity of available structures and difficulties in finding lead compounds that dock against flat, flexible protein-protein interaction surfaces. By employing computational docking of known small molecule inhibitors, we have demonstrated that structural ensembles derived from either accelerated MD (aMD) or MD in the presence of an organic cosolvent generally give better scores than those assessed from analogous conventional MD. Furthermore, conformations obtained from combined cosolvent aMD simulations started with the apo-Bcl-xL structure yielded better average and minimum docking scores for known binders than an ensemble of 72 experimental apo- and ligand-bound Bcl-xL structures. A detailed analysis of the simulated conformations indicates that the aMD effectively enhanced conformational sampling of the flexible helices flanking the main Bcl-xL binding groove, permitting the cosolvent acting as small ligands to penetrate more deeply into the binding pocket and shape ligand-bound conformations not evident in conventional simulations. We believe this approach could be useful for identifying inhibitors against other protein-protein interaction systems involving highly flexible binding sites, particularly for targets with less accumulated structural data.