Predicting Drug Resistance in Cancer Cell Lines Using Machine Learning Approaches

Cancers are the second leading cause of death worldwide, with breast cancer being the most prevalent type, accounting for 2.26 million cases in 2020. A major factor contributing to high cancer mortality rates is the ability of cancer cells to develop resistance to conventional therapies, highlighting the urgent need for further research and the development of more effective treatments. This prevalence led us to focus on predicting drug resistance in cancer cell lines, particularly breast cancer, using machine learning approaches. We propose a curated, comprehensive, and integrated dataset comprising gene expression profiles from the Cancer Cell Line Encyclopedia (CCLE), drug response data from the Genomics of Drug Sensitivity in Cancer (GDSC), and detailed drug information from PubChem. A notable advancement in our approach is the utilisation of the Morgan Fingerprint technique to effectively encode chemical compound structures to improve model compatibility and prediction outcomes. Furthermore, we employed the Deep Neural Network (DNN), which has demonstrated substantial improvements over existing models by achieving a root mean square error (RMSE) below 1. The results of our experiments underscore the model’s robust predictive capability and adaptability to new data.