SMARTDock: A Toolkit for the Automated Development of Target-Specific Scoring Functions Using Bioactivity Data.

Area of Science:

• Computational Chemistry
• Drug Discovery
• Bioinformatics

Background:

• Molecular docking is crucial for structure-based drug discovery, but scoring functions limit accuracy in virtual screening.
• Existing methods struggle to reliably rank active compounds from large libraries against biological targets.

Purpose of the Study:

• To introduce SMARTDock, a novel workflow enhancing virtual screening accuracy in drug discovery.
• To integrate machine learning and bioactivity data with GOLD docking for improved compound prioritization.

Main Methods:

• Developed SMARTDock, a platform-independent Dockerized workflow utilizing GOLD docking.
• Integrated protein-ligand interaction fingerprints (PADIF) and machine learning classification models.
• Leveraged publicly available bioactivity data (ChEMBL) for target-specific model training.

Main Results:

• SMARTDock successfully enhances virtual screening by improving the enrichment of active compounds.
• The PADIF-based machine learning methodology demonstrated improved screening performance across multiple targets.
• The workflow is user-friendly, accessible to computational and medicinal chemists, and requires minimal input data.

Conclusions:

• SMARTDock offers a significant advancement in structure-based drug discovery virtual screening.
• The integration of machine learning and bioactivity data provides a powerful tool for identifying drug candidates.
• This approach addresses limitations of traditional scoring functions, improving efficiency and accuracy.