Multimodal Machine Learning for Drug Knowledge Discovery

Abstract

Multimodal machine learning deals with building models that can process information from multiple modalities (i.e., ways of doing or experiencing something). Experiments involving humans are used to guarantee drug safety in the complex task of drug development. Drug-related data is readily available and comes in various modalities. The proposed study aims to develop novel methods for multimodal machine learning that can be used to process the diverse multimodal data used in drug development and other challenging tasks that could benefit from the use of multimodal data. We present a series of drug-related tasks which are used to both evaluate the models proposed in this ongoing study and discover new drug knowledge. This research will make far-reaching contributions to the field of machine learning, as well as practical contributions in the medical domain.