mycoCSM

mycoCSM predicts biological activity (minimum inhibitory concentration, MIC) and human safety of chemical compounds against Mycobacterium species, including Mycobacterium tuberculosis, using graph-based molecular signatures to prioritize candidates for anti-mycobacterial drug discovery.

Key Features:

• Graph-Based Signature Approach: Uses graph-based molecular signatures to encode structural information from molecular graphs for computational prediction of compound bioactivity.
• Training and Validation: Trained and validated on eight organism-specific datasets and a general Mycobacteria dataset with cross-validation correlation up to 0.89 and independent blind-test correlation of 0.88 for MIC predictions.
• Penetration Predictor: Includes a predictor for compound penetration into necrotic foci associated with tuberculosis, reporting a correlation coefficient of 0.75.
• Safety Estimation: Incorporates an estimator of human maximum tolerated dose to assess compound safety.

Scientific Applications:

• Screening Library Enrichment: Enriches screening libraries with potent anti-mycobacterial molecules.
• High-Throughput Activity and Safety Prediction: Provides high-throughput predictions of MIC and human safety to support early-stage drug discovery.
• Candidate Prioritization: Prioritizes compounds for further testing based on predicted activity, necrotic-foci penetration, and estimated maximum tolerated dose.

Methodology:

Employs graph-based molecular signatures trained and validated on eight organism-specific sets and a general Mycobacteria dataset, evaluated by cross-validation and independent blind tests using MIC as the activity endpoint, and includes models for necrotic-foci penetration and human maximum tolerated dose estimation.