GCFMCL: predicting miRNA-drug sensitivity using graph collaborative filtering and multi-view contrastive learning.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• MicroRNA (miRNA) mechanisms are crucial for drug action, impacting drug discovery and biomarker research.
• Traditional experimental methods for miRNA-drug susceptibility are costly and time-consuming.
• Existing deep learning methods struggle with sparse data and complex feature information in miRNA-drug interactions.

Purpose of the Study:

• To develop an efficient and accurate computational model for predicting miRNA-drug sensitivity.
• To address limitations in current deep learning approaches for analyzing miRNA-drug relationships.
• To leverage multi-view contrastive learning within a graph collaborative filtering framework.

Main Methods:

• Proposed GCFMCL, a graph collaborative filtering model incorporating multi-view contrastive learning.
• Developed topological contrastive learning using neighborhood information.
• Implemented feature contrastive learning to mine high-order feature correlations and neighborhood relationships.

Main Results:

• GCFMCL achieved high performance with AUC of 95.28%, AUPR of 95.66%, and F1-score of 89.77% on a dataset of 2049 miRNA-drug associations.
• The model significantly outperformed state-of-the-art methods.
• Multi-view contrastive learning effectively mitigated issues of sparse data and noise.

Conclusions:

• GCFMCL offers a powerful new approach for predicting miRNA-drug sensitivity.
• The model's performance demonstrates the effectiveness of multi-view contrastive learning in this domain.
• This work provides a valuable tool for drug target discovery and related research.