TIDAL: Topology-Inferred Drug Addiction Learning

Area of Science:

• Computational chemistry
• Artificial intelligence in drug discovery
• Pharmacology

Background:

• Drug addiction poses a significant global health challenge, necessitating efficient anti-addiction drug development.
• Traditional experimental drug screening is time-consuming and costly, highlighting the need for innovative computational approaches.

Purpose of the Study:

• To introduce Topology-Inferred Drug Addiction Learning (TIDAL), an AI framework for modeling and analyzing drug addiction data.
• To leverage advanced computational methods for identifying potential anti-addiction medications and predicting their properties.

Main Methods:

• Integration of multiscale topological Laplacians (embedding molecular topological and algebraic invariants) with a deep bidirectional transformer.
• Utilizing ensemble-assisted neural networks (EANNs) for enhanced predictive modeling.
• Validation on multiple datasets including drug addiction, hERG, and DAT.

Main Results:

• TIDAL demonstrates state-of-the-art performance in modeling and analyzing drug addiction data.
• Cross-target analysis identified drug-mediated linear and bilinear target correlations, revealing side effect and repurposing potentials.
• Application to existing anti-addiction medications provided insights into efficacy, repurposing, and side effects.

Conclusions:

• TIDAL offers a novel computational strategy for accelerating the development of anti-addiction medications.
• The framework aids in predicting drug efficacy, identifying repurposing opportunities, and alerting potential side effects.
• This AI-driven approach addresses the urgent need for efficient drug discovery in combating substance addiction.