Deep learning approaches to predict drug responses in cancer using a multi-omics approach

Abstract

Cancers are genetically heterogeneous, and therefore the same anti-cancer drug may have varying degrees of effectiveness on patients due to their different genetic profiles. Oftentimes, it is a trial and error process and patients have to try many different anti-cancers drugs that not are only ineffective, but also have significant side effects before finding one that is effective. The mechanisms of cancers are also an extremely complex, with many biological factors all contributing to their development, so we decided to take a multi-omics approach where we integrated multiple types of omics data in order to provide a more holistic molecular perspective on pharmacogenetics cancer research. With the development of deep learning, we have been able tackle the large amounts of complex omics data that is extremely challenging to process with conventional analytical methods, and the main objective of this project is to use deep learning to predict the response of tumours to different anticancer drugs using a multi-omics approach. So that doctors will be able to take a more customized approach to prescribe anti-cancer drugs that are likely to be more effective.