Target-specific support vector machine scoring in structure-based virtual screening: computational validation, in...

Area of Science:

• Computational chemistry and structural biology
• Drug discovery and development
• Bioinformatics and cheminformatics

Background:

• A previously developed structure-based support vector machine target-specific scoring function (SVM-SP) was evaluated across 41 targets.
• The Directory of Useful Decoys (DUD) dataset was utilized for comprehensive performance assessment.

Discussion:

• SVM-SP demonstrated superior enrichment performance compared to Glide and other scoring functions across diverse target families, especially kinases.
• The scoring function exhibited robust performance across different protein classes and even with homology models.
• High enrichment was achieved even with limited training data, indicating efficiency in model building.

Key Insights:

• Virtual screening of 1125 compounds against EGFR and CaMKII identified potent inhibitors.
• Three compounds showed significant in vitro inhibition of EGFR kinase activity and cancer cell proliferation.
• One compound demonstrated dose-dependent inhibition of CaMKII kinase activity.

Outlook:

• The findings suggest SVM-SP's potential for identifying chemical probes for kinases within the human kinome.
• This approach could accelerate drug discovery efforts targeting kinases, which are crucial in chemical biology.
• The success with non-kinase-like binding pockets indicates broader applicability beyond typical kinase inhibitors.