Grant: Discovery Grants Program - Individual

Our research program focuses on the development of novel methods to address outstanding questions in microbial genomics, with a particular emphasis on viruses, using data generated through long-standing collaborations, simulations, or that are publicly available. In the context of this long-term objective, this proposal will propel our program along 2 short-term objectives: (i) further develop methods tracing viruses from environmental sequencing data; (ii) developing machine learning (ML) approaches for predicting epistasis, protein-protein interactions, and epidemics from genomics data. This basic NSE program is purely computational, and does not involve any field work, sequencing, wet lab experiments or live microbes. The first short-term objective of our program aims to identify viruses in the environment, based sequencing data, and will test 2 hypotheses. First, we will develop tools to make specific identifications of viruses, and use evolutionary modeling to test how those are related to known viruses. Second, we will test how this relationship changes along latitudinal gradients, testing the hypothesis that undersampled environments, at high latitudes, contain viruses that are more and more "unknown." This second short-term objective is geared towards evolutionary and public health researchers, and is very timely, as a warming world is expected to lead to more viral spillovers and epidemics. Second, further developing recent work to detect co-evolution in proteins, we will test 2 hypotheses, gauging how ML can be used to (i) improve predictions, and (ii) detect actual epistasis. An innovative combination of auto-encoders, fitness models and Convolutional Neural Networks will be implemented, and tested on influenza and COVID-19 data. Then, we will test how ML can be used to predict viral outbreaks, bringing in two innovations: (i) real-time predictions ("nowcast predictions"), and (ii) genomics data to state-of-the-art models. This first short-term objective represents a fundamental axis of our program, aimed primarily at evolutionary biologists, with applications in outbreak surveillance and drug design. Altogether, these objectives will push our research program further by building innovative methods to not only increase our understanding of fundamental processes driving viral genomics and dynamics, but also to make available novel computational tools to the research community, with potential end-users ranging from evolutionary biologists, to population and health researchers. Knowledge will be disseminated through publications, conferences and seminars. Over the next 5 years, this program will train at least 5 graduate and more undergraduate HQP in state-of-the-art methods in computational biology, equipping them with highly transferable skills that are in high demand in the digital 21st century job market.