Mpropred

Mpropred predicts SARS-CoV-2 main protease (Mpro) inhibitory activity from chemical structures to prioritize compounds for drug discovery.

Key Features:

• Classification Structure-Activity Relationship (CSAR) model: Employs a CSAR model to identify substructures associated with anti-Mpro activity among 758 non-redundant compounds.
• Fingerprint descriptors: Utilizes a set of 12 molecular fingerprints to characterize Mpro inhibitors.
• Random Forest models: Implements a Random Forest algorithm to construct prediction models using 100 distinct data splits.
• Performance metrics: Evaluates model performance with accuracy (89%), sensitivity (89%), specificity (73%), Matthews correlation coefficient (79%), and a MODI index of 0.79.
• Descriptor analysis: Identifies structural features such as methyl side chains, aromatic rings, and halogen groups as significant for Mpro inhibition.

Scientific Applications:

• SARS-CoV-2 Mpro-targeted drug discovery: Prioritizes compounds for development against the SARS-CoV-2 main protease.
• High-throughput library screening: Facilitates rapid screening of compound libraries, exemplified by the CMNPD marine compound database, to identify candidate Mpro binders.
• Inhibitor evaluation and optimization: Integrates with molecular dynamics (MD) simulations and molecular mechanics/Poisson–Boltzmann surface area (MM/PBSA) analyses to evaluate and optimize potential inhibitors.

Methodology:

Computational methods explicitly include CSAR analysis of 758 non-redundant compounds, calculation of 12 molecular fingerprints, Random Forest model construction across 100 data splits, performance evaluation using accuracy, sensitivity, specificity, MCC and MODI, descriptor analysis of key substructures, and integration with MD and MM/PBSA analyses.