DPMGCDA: Deciphering circRNA-Drug Sensitivity Associations with Dual Perspective Learning and Path-Masked Graph...

Area of Science:

• Computational biology
• Genomics
• Pharmacology

Background:

• Circular RNAs (circRNAs) influence drug sensitivity and efficacy.
• Experimental validation of circRNA-drug associations is time-consuming and resource-intensive.

Purpose of the Study:

• To develop an efficient computational framework, DPMGCDA, for predicting circRNA-drug sensitivity associations.
• To overcome the limitations of traditional experimental methods.

Main Methods:

• Constructed circRNA-circRNA and drug-drug similarity networks using similarity network fusion.
• Built circRNA, drug, and circRNA-drug heterogeneous graphs.
• Employed dual perspective learning and a path-masked graph autoencoder for prediction.
• Integrated predictions from dual perspectives for final scoring.

Main Results:

• DPMGCDA significantly outperformed state-of-the-art methods in both transductive and inductive settings.
• Ablation tests confirmed the effectiveness of dual perspective learning.
• Embedding visualization demonstrated the path-masked graph autoencoder's feature encoding capabilities.
• Case studies validated DPMGCDA's ability to identify novel circRNA-drug associations.

Conclusions:

• DPMGCDA provides an effective computational approach for predicting circRNA-drug sensitivity associations.
• The method accelerates the discovery of potential biomarkers for drug efficacy.
• This framework aids in understanding the role of circRNAs in drug response.