Recommender system for cancer treatment

In cancer treatment, a patient's drug response is crucial as many patients do not respond to given drugs at all. Hence, an important goal is to precisely predict whether and especially why a patient will respond to certain treatments. In this thesis, the largest available cancer patients data set is used as a source to create response models for different types of cancer. With the help of machine learning, we show that with different kinds of data (clinical, gene or cell type fractions), it is possible to improve response predictions and gain potential insights on the causes.