Predicting Cancer Drug Response with Deep Learning

Predicting drug response in cancer cell lines is crucial for personalized medicine and drug development. This paper presents a deep learning approach that predicts pIC50 values using molecular fingerprints derived from SMILES representations.

Trained on the Genomics of Drug Sensitivity in Cancer (GDSC) dataset covering 985 cell lines and hundreds of compounds, our model achieves state-of-the-art prediction accuracy. The architecture combines molecular feature extraction with cell line genomic profiles to capture drug-target interactions.

We demonstrate the model's utility through an interactive tool that allows researchers to draw molecular structures and receive instant predictions across tissue types and cancer classifications.