MRNDR: Multihead Attention-Based Recommendation Network for Drug Repurposing.

Area of Science:

• Computational biology
• Pharmacology
• Artificial intelligence in drug discovery

Background:

• Drug development is costly and time-consuming.
• Drug repurposing offers a more efficient alternative but still requires extensive efficacy testing.
• Prescreening potential new indications for existing drugs can significantly reduce costs and accelerate the process.

Purpose of the Study:

• To introduce a novel drug repurposing recommendation model, MRNDR (Multi-head attention-based Recommendation Network for Drug Repurposing).
• To develop a predictive tool for drug-disease relationships that enhances the efficiency of drug repurposing.
• To leverage advanced AI techniques for accurate and cost-effective drug repurposing.

Main Methods:

• Utilized a million-level training dataset (BioRE - Biology Recommended Entity data).
• Implemented a multi-head self-attention mechanism for robust generalization.
• Employed the proposed Weighted Representation Distance Score (WRDS) algorithm.

Main Results:

• Achieved state-of-the-art performance on the GP-KG public dataset.
• Obtained an MRR (Mean Reciprocal Rank) of 0.308 and a Hits@10 score of 0.628.
• Demonstrated significant improvements of 4.7% (MRR) and 18.1% (Hits@10) over existing models.

Conclusions:

• The MRNDR model effectively predicts drug-disease relationships, facilitating efficient drug repurposing.
• Validation through clinical trial data indirectly confirms the practical applicability of MRNDR's recommendations.
• MRNDR reduces the need for manual expert assessment, streamlining the drug repurposing pipeline.