DeepHeteroCDA: circRNA-drug sensitivity associations prediction via multi-scale heterogeneous network and graph...

Area of Science:

• Computational biology
• Genomics
• Pharmacology

Background:

• Drug sensitivity is critical for effective cancer treatment and understanding drug resistance.
• Circular RNAs (circRNAs) are emerging biomarkers with potential in disease research and therapeutics.
• Identifying associations between circRNAs and cellular drug sensitivity is key to advancing precision medicine.

Purpose of the Study:

• To develop a novel computational method, DeepHeteroCDA, for predicting associations between circRNAs and cellular drug sensitivity.
• To leverage multi-scale heterogeneous networks and graph attention mechanisms for accurate prediction.
• To provide a tool for understanding circRNA-mediated drug response and resistance.

Main Methods:

• Constructed a heterogeneous graph integrating drug-drug similarity, circRNA-circRNA similarity, and known circRNA-drug sensitivity associations.
• Employed graph convolutional networks (GCN) to embed drug 2D structures within the heterogeneous graph.
• Utilized a graph attention network and GCN within a multi-scale heterogeneous network framework for association prediction.

Main Results:

• DeepHeteroCDA demonstrated superior performance in predicting circRNA-drug sensitivity associations compared to existing methods.
• Visualization experiments confirmed the model's ability to effectively extract relevant association information.
• Case studies validated the model's effectiveness in identifying potential and novel circRNA-drug sensitivity links.

Conclusions:

• DeepHeteroCDA offers a powerful and effective approach for predicting circRNA-drug sensitivity associations.
• The findings contribute to a deeper understanding of circRNA roles in drug response and resistance.
• The developed method holds promise for advancing personalized cancer therapy and drug development.