Hsp90 inhibitors, part 2: combining ligand-based and structure-based approaches for virtual screening application

Abstract

Hsp90 continues to be an important target for pharmaceutical discovery. In this project, virtual screening (VS) for novel Hsp90 inhibitors was performed using a combination of Autodock and Surflex-Sim (LB) scoring functions with the predictive ability of 3-D QSAR models, previously generated with the 3-D QSAutogrid/R procedure. Extensive validation of both structure-based (SB) and ligand-based (LB), through realignments and cross-alignments, allowed the definition of LB and SB alignment rules. The mixed LB/SB protocol was applied to virtually screen potential Hsp90 inhibitors from the NCI Diversity Set composed of 1785 compounds. A selected ensemble of 80 compounds were biologically tested. Among these molecules, preliminary data yielded four derivatives exhibiting IC50 values ranging between 18 and 63 μM as hits for a subsequent medicinal chemistry optimization procedure.

Figure 1

Overview of (A) the applied procedure and (B) alignment assessment protocol.

Figure 2

Predicted activity contribution plots (solid: 75%, positive: green, negative: yellow), overlapped with PLS-coefficients plots (mesh: 65%, positive: red, negative: blue) obtained from the used 3-D QSAR models at the selected PC, for the most active screened compounds in their BC system (protein and pose): NCI23128 in 1UY6, NCI610930 in 1UYC, NCI117285 and NCI170578 in 1UY8.

Figure 3

The most active screened compounds in their BC system (protein and pose) overlapped with 2YK2(16) (Hit 1 in orange, Hit 2 in cyan, protein in cyan): (A) NCI610930 in 1UYC (ligand and protein in green), (B) NCI117285 in 1UY8 (ligand in yellow, protein in magenta), and (C) NCI170578 in 1UY8 (ligand and protein in magenta).

Figure 4

Depiction of the most active screened compound NCI610930 in its BC system (ligand and protein in green) overlapped with 2YKI(16) (brown colored).