Optimizing drug synergy prediction through categorical embeddings in Deep Neural Networks

Abstract

AbstractCancer treatments often lose effectiveness as tumors develop resistance to single-agent therapies. This challenge can be addressed through combinatorial treatments, where multiple drugs are administered simultaneously. However, the combinatorial space ofNdifferent drugs atddoses results inNdpossible treatments, making it impractical to test each combination experimentally. To efficiently identify the most promising drug combinations, we propose a data-driven approach utilizing deep learning with categorical embeddings. Our method leverages neural networks to extract meaningful patterns from categorical variables, optimizing the prediction of drug synergy and enhancing treatment efficacy.