Towards Effective Consensus Scoring in Structure-Based Virtual Screening.

Area of Science:

• Computational chemistry
• Drug discovery
• Bioinformatics

Background:

• Virtual screening (VS) is crucial for identifying drug candidates but is often limited by data and method specificity.
• Conventional VS methods rely on physicochemical descriptors, leading to variable success rates in drug candidate identification.
• Existing docking programs exhibit differing utilities, necessitating more robust and generalizable screening approaches.

Purpose of the Study:

• To develop a general virtual screening (VS) formulation using novel consensus scoring (CS) algorithms.
• To enhance the accuracy and reliability of identifying active ligands for diverse protein targets.
• To provide a parsimonious and computationally efficient alternative to existing VS methods.

Main Methods:

• Developed four universality classes of consensus scoring (CS) algorithms.
• Integrated docking scores from ten distinct docking programs (ADFR, DOCK, Gemdock, Ledock, PLANTS, PSOVina, QuickVina2, Smina, Autodock Vina, VinaXB).
• Utilized the DUD-E decoy repository against 29 MRSA-oriented targets for validation.

Main Results:

• Consensus scoring (CS) demonstrated improved ligand-protein docking fidelity compared to individual docking platforms.
• The CS approach requires a minimal number of docking program combinations for effective screening.
• CS algorithm predictions showed strong agreement with independent machine learning evaluations.

Conclusions:

• Consensus scoring (CS) offers a reliable and generalizable virtual screening (VS) formulation for drug discovery.
• This method effectively identifies high-affinity ligands and suitable protein targets for drug repositioning.
• CS provides a computationally efficient and accurate alternative to traditional, resource-intensive docking approaches.