Exploring ncRNA-Drug Sensitivity Associations via Graph Contrastive Learning

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Noncoding RNAs (ncRNAs) significantly influence drug efficacy by regulating genes associated with drug sensitivity.
• Identifying ncRNA-drug sensitivity associations is vital for advancing drug discovery and disease prevention strategies.
• Traditional experimental methods for this identification are often inefficient, time-consuming, and labor-intensive.

Purpose of the Study:

• To develop and validate a novel computational approach, NDSGCL, for predicting ncRNA-drug sensitivity associations.
• To leverage graph contrastive learning for enhanced feature representation of ncRNAs and drugs within bipartite graphs.
• To improve the efficiency and accuracy of identifying ncRNA-drug interactions compared to existing methods.

Main Methods:

• Developed NDSGCL, a graph contrastive learning framework utilizing graph convolutional networks.
• Integrated local structural neighbors and global semantic neighbors for comprehensive feature learning in ncRNA-drug bipartite graphs.
• Employed contrastive learning objectives to capture higher-order relationships and mitigate data sparsity.

Main Results:

• NDSGCL demonstrated superior performance over baseline graph convolutional network methods, existing contrastive learning approaches, and state-of-the-art prediction models.
• Visualization experiments confirmed the significant contribution of both local structural and global semantic contrastive objectives.
• Case studies involving two specific drugs validated NDSGCL's effectiveness in predicting ncRNA-drug sensitivity associations.

Conclusions:

• NDSGCL offers a powerful and efficient computational tool for predicting ncRNA-drug sensitivity.
• The integration of local and global contrastive learning strategies significantly enhances prediction accuracy.
• This approach holds promise for accelerating drug discovery and personalized medicine by elucidating ncRNA-drug interactions.