H$^{3}$CDR : An Anti-Cancer Drug Response Prediction Model Driven by Heterogeneous and Homogeneous Hybrid Graph...

Area of Science:

• Computational biology
• Bioinformatics
• Genomics

Background:

• Cancer treatment efficacy varies due to disease heterogeneity.
• Predicting drug response is crucial for personalized cancer therapy.
• Current methods struggle to capture comprehensive cell line and drug features.

Purpose of the Study:

• To develop an advanced framework for predicting anticancer drug response (CDR).
• To address limitations in existing CDR prediction models by integrating homogeneous and heterogeneous features.
• To enhance precision oncology through improved therapeutic effect prediction.

Main Methods:

• Proposed a hybrid graph neural network framework named H3CDR.
• Fused multi-omics data to learn similarity features between cancer cell lines and drugs.
• Employed a multi-branch network for comprehensive feature extraction from both cell lines and drugs.

Main Results:

• H3CDR demonstrated superior performance compared to existing methods.
• Achieved an Area Under the ROC Curve (AUC) of 0.8772.
• Achieved an Area Under the Precision-Recall Curve (AUPRC) of 0.8819 on the GDSC dataset.

Conclusions:

• H3CDR effectively predicts anticancer drug response by integrating multi-omics data and advanced network architecture.
• The model enhances the prediction of therapeutic effects, advancing precision oncology.
• The framework offers a promising approach for personalized cancer treatment strategies.