Simultaneous regression and classification for drug sensitivity prediction using an advanced random forest method

Abstract

Machine learning methods trained on cancer cell line panels are intensively studied for the prediction of optimal anti-cancer therapies. While classifcation approaches distinguish efective from inefective drugs, regression approaches aim to quantify the degree of drug efectiveness. However, the high specifcity of most anti-cancer drugs induces a skewed distribution of drug response values in favor of the more drug-resistant cell lines, negatively afecting the classifcation performance (class imbalance) and regression performance (regression imbalance) for the sensitive cell lines. Here, we present a novel approach called SimultAneoUs Regression and classifcatiON Random Forests (SAURON-RF) based on the idea of performing a joint regression and classifcation analysis. We demonstrate that SAURON-RF improves the classifcation and regression performance for the sensitive cell lines at the expense of a moderate loss for the resistant ones. Furthermore, our results show that simultaneous classifcation and regression can be superior to regression or classifcation alone.