AI models to predict response to drug combinations in poor outcome cancer patients

Principal Investigator: Amin Emad, Morag Park
Theme : Artificial Intelligence (AI) / Health
Competition : Omics Data Against Cancer (ODAC) Competition
Status : Completed
Start : Oct. 1, 2020
End: Sept. 30, 2022
Budget : $300,000.00


PROJECT COLLABORATION WITH IVADO AND ONCOPOLE

 

Cancer is a major health concern in Canada and Quebec. Precision medicines have the potential to improve patient prognosis by matching treatments to the characteristics of a patient’s disease. However, many cancers lack precision medicine therapies and only receive the “standard of care”, usually toxic chemotherapies for their cancer type, many of which have a low response rate and strong side-effects. For example, a certain type of cancer called triple negative breast cancer, which accounts for ~15% of all breast cancer, is the most aggressive subtype of breast cancer and lacks targeted therapies. In addition, many patients with an initial response to chemotherapies develop resistance followed by progressive disease and relapse.

 

 

One promising approach to improve patients’ prognosis and overcome drug resistance is the use of combinational therapies (administering multiple drugs with synergistic effects simultaneously). However, predicting which combinational therapies (out of tens of thousands potential candidates) should be used for an individual is a major problem. One reason is that even if two drugs have synergistic effects in in vitro cell line experiments, a patient may still show resistance to that combination due to the unique molecular and clinical characteristics of their disease. While traditionally factors such as cancer types or symptoms were used to identify treatment options, molecular “omics” profiling of tumours (e.g. expression of tens of thousands of genes) now provides us with an abundance of data to address this challenge using artificial intelligence (AI). Particularly, advanced “deep learning” approaches that have the ability to model complex nonlinear relationships among molecular features and drug response have the potential to revolutionize precision medicine.   

 

 

In this project, we propose to develop such deep learning models to enable the prediction of the drug response of cancer patients to precision combinational therapies using their available molecular “omics” and clinical characteristics. The proposed computational models developed here, which will be made accessible to the community, will result in a versatile “in-silico clinical trial” framework that can be tested by us experimentally. Such framework will provide an opportunity to determine the best course of treatment for a patient and suggest novel drug combinations to treat and/or overcome cancer drug resistance, ultimately impacting the lives of millions around in Canada and around the world.

 

Lead Genome Centre: Génome Québec 

Partners: IVADO, Oncopole