CLC-Pred 2.0

CLC-Pred 2.0 predicts cytotoxic effects of chemical and natural compounds on normal and cancer human cell lines from structural formulas.

Key Features:

• Dataset Coverage: Utilizes 128,545 structures from ChEMBL and PubChem to support cytotoxicity prediction across 391 tumor and 47 normal human cell lines.
• NCI60 Panel: Incorporates NCI60 data comprising 22,726 structures to predict cytotoxicity with IC50 thresholds of 100, 10, and 1 nM.
• Predictive Accuracy (tumor and normal): Reports mean AUCs of 0.925 (LOO CV) and 0.923 (20F CV) for tumor and normal cell line predictions.
• NCI60 Predictive Range: Reports AUC ranges of 0.870–0.945 (LOO CV) and 0.869–0.942 (20F CV) for the NCI60 panel.
• Mechanism of Action Prediction: Predicts 2,170 molecular mechanisms of action with AUCs of 0.979 (LOO CV) and 0.978 (20F CV) based on ChEMBL and PubChem data.
• Validation Protocols: Uses leave-one-out (LOO CV) and 20-fold cross-validation (20F CV) to report predictive performance metrics.

Scientific Applications:

• Drug Discovery and Development: Prioritizes compounds for antineoplastic screening by providing predicted cytotoxicity profiles across tumor cell lines.
• Safety Evaluation: Assesses potential cytotoxic effects on normal human cell lines to support early safety profiling of compounds.
• Mechanistic Insights: Provides predicted molecular mechanisms of action to inform hypotheses about compound–target or pathway interactions.

Methodology:

Employs the Prediction of Activity Spectra for Substances (PASS) approach using substructural atom-centric MNA descriptors and integrates a Bayesian algorithm.