Predicting strain interactions and success with genomic data: approaches from modelling and machine learning
SPEAKER: Caroline Colijn
Professor and Canada 150 Research Chair
Department of Mathematics
Simon Fraser University
ABSTRACT: Genomic data gathered over time give the opportunity to examine diversity at high resolution, and to use the results to test predictive models: can we guess in advance which sub-populations of a circulating pathogen will succeed in the near future? We discuss two contrasting approaches, one based on machine learning and the other using a high-dimensional mechanistic model that empirically captures features of the genomic data. We illustrate these approaches using data from human influenza A, Staphylococcus aureus and Streptococcus pneumoniae.