RF QSAR

RF QSAR predicts ligand-target interactions using 1121 structure-activity relationship (SAR) models and transforms model scores into probabilities to estimate ligand-target interaction likelihoods.

Key Features:

• Model Composition: 1121 SAR models built with the random forest algorithm for robust predictions.
• Target Ranking: Achieves 67.6% and 73.9% recall rates for top-11 and top-33 targets.
• Unified Scoring: Transforms model scores into probabilities to estimate ligand-target interaction likelihood.

Scientific Applications:

• Drug Discovery: Identifies potential targets for novel compounds and elucidates polypharmacology.

Methodology:

Constructs SAR models using random forest, validates via five-fold cross-validation and external datasets, and employs a unified scoring scheme.