Structure-based Virtual Screening In Lead Compound Discovery

Structure-based virtual screening (molecular docking) plays important role indiscovering novel lead compounds. To improve the accuracy of ligand bindingfree-energy calculation, we systematically investigated physics-based MM-PB/SAmethod based on previous published work. This thesis mainly includes the followingaspects:1. Current default MM-PB/SA protocol may be improved, including the setup ofsolvent model and the adjustments of parameters. We demonstrated that the optimizedMM-PB/SA protocol significantly improves the prediction power in data setcontaining three representative drug targets.2. To impletement MM-PB/SA method in the practice of lead compounddiscovery, we developed a hierarchical virtual screening strategy to strike thebanlance between efficiency and accuracy. We evaluated the performance of thisstrategy in the case of tubulin. Our hierarchical strategy generated63candidatesautomatically from more than100000diverse drug-like compounds.9molecules withsimilar strcture and predicted binding modes were selected for experimental test,among which5were active. Further structure-activity relationship study validated ourcomputational model.3. We further developed combinatorial virtual screening method to exploremulti-target inhibitors. Using BCR-ABL protein as our model system, we aimed toidentify dual-inhibitors to target the ATP-site on both wild-type ABL kinase anddrug-resistance T315I mutant. Consequently,3of9candidates resulted from virtualscreening were experimentally validated as true binders with significantdual-inhibition activity. Encouragingly, the identified candidates inhibit proliferationof ABL-driven Ba/F3cell lines (both wild-type and drug-resistance) comparing withnormal Ba/F3cell line. These compounds also show inhibition on BCR-ABL highlyexpressed tumor cell, such as K562.4. Interestingly, one same lead compound was identified to target both tubulinand ABL kinase. Thus, we discovered the similar binding characteristics between twocompletely dissimilar binding sites, colchicine site in Tubulin and ATP binding site inABL kinase. Importantly, these characteristics can’t be revealed by conventionalsequence or structural similarity analysis.