ASCoVPred

ASCoVPred predicts anti-SARS-CoV-2 activity and human cell toxicity of molecular compounds using quantitative structure-activity relationship (QSAR) modeling and machine learning (ML) to prioritize candidate therapeutics.

Key Features:

• QSAR modeling: Uses quantitative structure-activity relationship models to relate molecular descriptors and fingerprints to biological activity.
• Descriptor and fingerprint calculation: Computes compound descriptors and fingerprints using PaDEL v2.21 for input feature generation.
• Machine learning implementation: Employs ML algorithms implemented in Weka v3.8.2 to build predictive models.
• Training data: Models are trained and optimized with experimentally validated SARS-CoV-2 inhibitory compounds.
• Feature selection: Applies rigorous feature selection to identify decisive molecular descriptors and fingerprints correlated with activity and toxicity.
• Predictive outputs: Provides predictions of anti-SARS-CoV-2 activity and minimal human cell toxicity for compound evaluation.

Scientific Applications:

• Compound discovery: Identification and prioritization of potential anti-SARS-CoV-2 compounds based on predicted activity and toxicity.
• Lead optimization: Guiding refinement of molecular candidates by highlighting descriptors associated with improved activity and reduced toxicity.
• Preclinical screening support: Filtering and ranking compounds for experimental validation against SARS-CoV-2.

Methodology:

QSAR modeling using PaDEL v2.21-derived descriptors and fingerprints as features, machine learning algorithms implemented in Weka v3.8.2, training and optimization on experimentally validated SARS-CoV-2 inhibitory compounds, and rigorous feature selection to identify decisive descriptors and fingerprints.