Machine learning model for breast anticancer drug sensitivity prediction from gene expression

Therapeutic action of drugs and their potential mechanisms provide an important basis for precision medicine. Cancer cell lines and drug sensitivities associated with different compounds create a valuable source for researchers to study the therapy response and help to convert in vitro findings of cell lines into in vivo therapeutic designs which will be reflected on patient care. In this study, we trained a predictive model for 26 anticancer drugs. These were chosen as the most used compounds in breast cancer therapy. These were aligned with gene expression data correlated with drug sensitivity measured by drugs' efficacy IC50 of cancer cell lines for different tumor tissues. The study identified drug gene associations using datasets from Cancer Cell Line Encyclopedia (CCLE). This research built a model that can predict drug resistance based on the sensitivity value IC50. Through this study we managed to predict which drug compound class is suitable for which cancer tissue. Also, we found that according to genetic expression we can predict more cancer suppressing drugs. This model can be used for predicting preclinical drug trials for effectiveness across breast cancer types. Our proposed model successfully achieved an accuracy of over 70% of the 26 selected drug compounds.