The MIDAS Webinar Series features research by MIDAS members, and is open to the public.
Date: Friday, November 17, 2023
Speaker: The Ruian Ke Lab
Abstract:
SARS-CoV-2 is continuously evolving new variants causing waves of infection globally, leading to large numbers of infections and a high death toll. With dozens or hundreds of minor variants circulating in the global population, there is an urgent need for predicting the scale and the rate of the spread of a new variant when it emerged in the population. This would allow for more focused experimental efforts and for timely formulation of new vaccines.
Here in this talk, I will present two of our ongoing works that use machine learning and graph theory approaches to make such predictions based on early viral genomic data. In the first work, we employed a graph theory approach to analyze the collection of pairwise distance matrices derived from genetic sequences of SARS-CoV-2 lineages. Extracting features from these matrices, we derived a simple model to accurately predict the rate of spread of a newly emerged lineage using early available sequences. The model is able to identify the future-dominating lineages when their lineage frequencies were still low (e.g. <5%). In the second work, we constructed a machine learning model, based on the transformer architecture (used in modern language processing models), and trained the model using existing lineage frequency time series. We show that this model could predict the frequency of a newly emerged lineage 2 months into the future with a high level of accuracy. Overall, these two approaches represent promising new methods utilizing genomic data for SARS-CoV-2 lineage monitoring and forecasting.
Bio:
Ruian Ke is a staff scientist at the T6: Theoretical Biology and Biophysics Group at Los Alamos National Laboratory (LANL). Before joining LANL, he was an Assistant Professor of Mathematics and Precision Medicine at North Carolina State University between 2015 and 2018. His recent research has particularly focused on developing data science approaches and machine learning tools to address key questions in the infection and transmission dynamics of SARS-CoV-2.