Scalable Permutation Invariant Multi-Output Gaussian Processes for Cancer Drug Response

Dose-response prediction in cancer is a critical step to assessing the efficacy of drug combinations on cancer cell-lines. The efficacy of a pair of drugs can be expressively modelled through a dose-response surface which outputs the viability score across a spectrum of drug concentrations for each pair of drugs in the training data. Using large in-vitro drug sensitivity screens, the goal is to develop accurate predictive models that can be used to inform treatment decisions by predicting the efficacy of given drug combination on new cancer cell lines as well as predict the effect of unseen drugs. Previous work, proposed a framework for modelling dose response surfaces with multi-output GPs, however, the model relied on the exact GP marginal likelihood and prohibited scalable inference. Further, the only inputs were drug concentrations per pair while the triplet of cell-lines and drug pair corresponded to different outputs . We make two important innovations in this work, we propose a framework for stochastic multi-output GPs for scalable inference; and, use a deep generative model (DGM) to embed the drugs in a continuous chemical space - enabling viability predictions for unseen drugs. We demonstrate the performance of our model in a simple setting using a high-throughput dataset and show that the model is able to efficiently borrow information across outputs.