COMET: A Machine-Learning Framework Integrating Ligand-Based and Target-Based Algorithms for Elucidating Drug...

Area of Science:

• Computational chemistry
• Drug discovery
• Bioinformatics

Background:

• Identifying molecular targets for bioactive compounds (target-fishing) is crucial in drug discovery.
• Computational methods accelerate target identification, reducing experimental workload.
• Existing methods require improvement in accuracy and efficiency.

Purpose of the Study:

• To develop a novel computational target-fishing method named COMET.
• To integrate ligand-based and target-based scoring for enhanced target ranking.
• To provide a user-friendly and efficient tool for drug discovery researchers.

Main Methods:

• COMET utilizes a random forest algorithm integrating ligand similarity and target binding scores.
• It leverages curated datasets of 2685 human targets and 990,944 ligand-target interactions.
• A modular framework allows for future method enhancements.

Main Results:

• COMET demonstrated superior performance compared to seven other methods in head-to-head comparisons.
• Achieved consistent top 100 recalls and 70-80% top 15 hit rates on large test sets.
• Showcased an optimal balance between computational speed and prediction accuracy.

Conclusions:

• COMET is a highly accurate and efficient computational tool for drug discovery target-fishing.
• Its performance and speed make it suitable for practical applications in pharmaceutical research.
• The tool is available for free trials, facilitating wider adoption and validation.