Predicting breast cancer drug response using a multiple-layer cell line drug response network model

Area of Science:

• Computational Biology
• Genomics
• Pharmacogenomics

Background:

• Precision medicine in breast cancer (BC) aims to predict patient drug response using molecular profiles.
• Existing models often lack validation on patient-derived data, do not consider drug properties, and focus on single-gene analysis.
• Gene expression profiles are highly informative for drug response prediction but often underutilized in integrated models.

Purpose of the Study:

• To develop an advanced drug response prediction model for breast cancer (BC).
• To integrate multiple data types, including cell line and drug information, for improved prediction accuracy.
• To validate the model's performance on BC cell lines and patient-derived samples.

Main Methods:

• Collected gene expression profiles and IC50 values for 49 BC cell lines and 220 drugs from GDSC.
• Developed a multiple-layer cell line-drug response network (ML-CDN2) integrating cell line and drug similarity networks.
• Utilized ML-CDN2 to predict drug responses in new BC cell lines and patient-derived samples.

Main Results:

• ML-CDN2 achieved a high predictive performance (Pearson correlation coefficient of 0.873) on GDSC cell line-drug pairs.
• The model demonstrated good predictive accuracy (Pearson correlation coefficient of 0.718) on external BC cell lines from CCLE.
• ML-CDN2 successfully predicted drug response in breast cancer patient-derived samples from TCGA.

Conclusions:

• The ML-CDN2 model effectively predicts breast cancer drug response by integrating comprehensive cell line and drug data.
• ML-CDN2 shows significant potential for predicting drug response in patient-derived samples, outperforming existing methods.
• This model represents a step forward in developing personalized treatment strategies for breast cancer patients.