Virtual drug screen schema based on multiview similarity integration and ranking aggregation

Area of Science:

• Computational chemistry
• Drug discovery
• Bioinformatics

Background:

• Current virtual screening (VS) methods, ligand/target structure-based and protein-ligand interaction fingerprint-based, offer complementary information.
• Integrating these diverse VS approaches presents a challenge for comprehensive evaluation and efficient drug screening.

Purpose of the Study:

• To develop and evaluate an efficient virtual screening schema integrating multiple VS methods from a comprehensive multiview perspective.
• To demonstrate the benefits of multiview similarity integration and ranking aggregation for drug VS.

Main Methods:

• A virtual screening schema was developed using multiview similarity integration and ranking aggregation.
• The schema was tested using HIV-1 protease complex structures and connectivity map data with various drug representations and fingerprints.
• Ritonavir, HDAC, and HSP90 inhibitors were used as query molecules.

Main Results:

• Rank aggregation significantly enhanced similarity searching results when multiple descriptions were involved.
• Integrated VS based on multiple data fusion outperformed individual VS methods.
• The study provides a promising approach for efficient drug screening using heterogeneous data sources.

Conclusions:

• Multiview integration and ranking aggregation are effective strategies for improving virtual screening performance.
• Combining diverse data sources and computational approaches leads to more robust and accurate drug candidate identification.
• This integrated VS approach offers a significant advantage for modern drug discovery.