Integration of autoencoder and graph convolutional network for predicting breast cancer drug response

Area of Science:

• Computational biology
• Genomics
• Pharmacology

Background:

• Breast cancer is the most common cancer in women, with treatment efficacy often hindered by tumor heterogeneity.
• Understanding and predicting patient response to anticancer drugs is crucial for effective treatment design.
• Integrating multi-omics and drug structure data can enhance the performance of cancer drug response prediction models.

Purpose of the Study:

• To develop and evaluate a novel computational model for predicting anticancer drug response in breast cancer.
• To investigate the utility of integrating multi-omics data with drug structure information for improved prediction accuracy.
• To classify cell lines as sensitive or resistant to specific drugs based on predicted responses.

Main Methods:

• Proposed an Autoencoder (AE) and Graph Convolutional Network (AGCN) model for drug response prediction.
• Utilized AE for dimensionality reduction of high-dimensional omics data (genetic, transcriptomic).
• Integrated reduced omics features with drug structure data via AGCN and a Convolutional Neural Network (CNN) to predict IC50 values, followed by K-means clustering for sensitivity classification.

Main Results:

• The AGCN model achieved an accuracy of 0.82 in predicting drug responses.
• External validation using The Cancer Genome Atlas (TCGA) data yielded a higher accuracy of 0.91.
• The results demonstrate superior performance compared to existing methods.

Conclusions:

• Integrating multi-omics data with drug structure data using AGCN significantly enhances drug response prediction accuracy.
• The developed model shows promise for personalized medicine in breast cancer treatment.
• Accurate drug response prediction is vital for advancing cancer therapy and drug design.