Improved and optimized drug repurposing for the SARS-CoV-2 pandemic

Area of Science:

• Computational biology
• Drug discovery
• Artificial intelligence in medicine

Background:

• The SARS-CoV-2 pandemic necessitates rapid development of effective treatments.
• De novo drug development is time-intensive; drug repurposing offers a faster alternative.
• Knowledge graphs and Graph Neural Networks (GNNs) are effective for drug repurposing.

Purpose of the Study:

• To extend the Dr-COVID model for enhanced drug repurposing in COVID-19.
• To improve the interpretation and application of GNN-based drug repurposing predictions.
• To optimize computational efficiency for large-scale drug candidate identification.

Main Methods:

• Utilized advanced Graph Neural Networks (GNNs) and knowledge graphs.
• Implemented novel output interpretation and aggregation strategies for drug ranking.
• Applied collaborative filtering for generating additional drug candidates.
• Optimized model implementation using data-parallelism and Non-Volatile Memory.

Main Results:

• Generated a top-100 ranking of 8,070 potential drug candidates.
• Identified 32 drugs from the ranking that are currently in COVID-19 clinical trials.
• Demonstrated significant reductions in inference and pre-processing times.
• Showcased accelerated post-processing using high-capacity Non-Volatile Memory.

Conclusions:

• The enhanced Dr-COVID model significantly improves drug repurposing efficiency and candidate identification for COVID-19.
• The study validates the potential of AI-driven approaches in accelerating therapeutic discovery.
• Integration of advanced hardware and computational strategies is crucial for large-scale drug repurposing efforts.