Computational Drug Repurposing Based on a Recommendation System and Drug-Drug Functional Pathway Similarity

Area of Science:

• Oncology
• Pharmacology
• Bioinformatics

Background:

• Drug repurposing offers a faster route to new cancer therapies.
• Cancer's heterogeneity and resistance necessitate novel treatment strategies.
• Existing drug repurposing methods have limitations in capturing complex drug response mechanisms.

Purpose of the Study:

• To develop and validate a computational recommendation system for prioritizing candidate cancer drugs.
• To create a drug-drug pathway functional similarity metric integrating multi-omics data.
• To identify synergistic drug combinations for cancer treatment.

Main Methods:

• Utilized a recommendation system learning model for drug prioritization.
• Designed a drug-drug pathway functional similarity by integrating gene expression, copy number variation (CNV), and DNA methylation data.
• Compared the proposed method with similarities based on chemical structures (SMILES) and protein-protein interaction networks.

Main Results:

• The multi-omics pathway similarity approach yielded more interpretable models of drug response.
• The method achieved accuracy comparable to other learning models on large public datasets (CCLE, GDSC).
• A case study demonstrated a synergistic effect of Erlotinib and OSI-906 (Linsitinib) in reducing tumor growth.

Conclusions:

• The computational method effectively characterizes drug response through multi-omics pathways.
• Systematic prediction of candidate cancer drugs with similar therapeutic effects is achievable.
• The findings suggest a promising alternative targeted therapy approach for cancer patients.