CoREx

CoREx applies machine learning and network-medicine analyses to prioritize FDA-approved drugs by their predicted interactions with protein targets, host interactomes, and protein functions relevant to SARS-CoV-2.

Key Features:

• Matrix factorization ranking: Matrix factorization algorithms rank broad-spectrum antivirals by predicted effectiveness and identified candidates overlapping with drugs used under compassionate use for COVID-19.
• Graph-kernel network analysis: Graph kernels evaluate drug-induced perturbations within a SARS-CoV-2–relevant subnetwork of the human interactome to prioritize compounds.
• Data integration: Integration of biological networks, protein functions, clinical drug usage, and Connectivity Map perturbation data supports comparative analyses.
• Visualization of molecular relationships: Visualization of relationships among drugs, protein targets, interactomes, and protein functions facilitates interpretation of predicted interactions.
• Experimental concordance: Predictions from the graph-kernel method show concordance with available experimental data.

Scientific Applications:

• Drug repositioning for SARS-CoV-2: Prioritizing FDA-approved drugs for potential repurposing against SARS-CoV-2.
• Antiviral ranking: Ranking broad-spectrum antivirals by predicted effectiveness for further experimental evaluation.
• Network-medicine analysis: Assessing perturbations in human interactome subnetworks relevant to viral infection and replication.
• Perturbation signature comparison: Leveraging Connectivity Map signatures to compare and interpret drug-induced molecular responses and support hypothesis generation.

Methodology:

CoREx implements two machine-learning approaches—matrix factorization algorithms to rank antivirals and graph-kernel–based network-medicine analyses to evaluate drug-induced perturbations within a SARS-CoV-2–relevant subnetwork of the human interactome—while integrating biological networks, protein functions, clinical drug usage, and Connectivity Map data.