Designing molecules with desired properties is a fundamental problem in drug, vaccine, and material discovery. Traditional approaches to designing new drugs can take over 10 years and a billion US dollars. Artificial intelligence (AI) can revolutionize molecule discovery by analyzing evidence from large amounts of data accumulated and learning how to search in the compositional space of molecules, and hence significantly accelerate and improve the process.
In this project, researchers will focus on two biological application challenges and novel AI algorithms with two methodological objectives. The interrelated biological application challenges are, on the one hand, the discovery of new antimicrobials, more specifically those based on antimicrobial peptides, and, on the other hand, active learning methods to better ascertain the relevant gene regulatory network, seen as a causal explanation of the effect of antibiotics. On the methodological side, researchers propose advances on methods to better model the acquired data, in particular introducing the ability to estimate the epistemic uncertainty (the reducible modeling ambiguity due to lack of data) over biological causal mechanisms. Researchers also propose advances in experimental design methods based on these posteriors to better estimate the information gain expected from a candidate experiment, so as to better design subsequent experimental cycles and reduce the time needed to discover a drug. This multidisciplinary project raises exciting fundamental challenges in AI, biology, genomics and peptide chemistry that will be tackled by gathering expertise from Mila, University de Montreal and McGill researchers, with many further applications to scientific discovery more broadly.
This project is part of the IVADO strategic funding program : AI for the discovery of materials and molecules
Lead Genome Centre: Génome Québec
Partner: IVADO