Professor Magnus Rattray (University of Manchester) will deliver a seminar on his work on how to learn models and make inferences given evidence from high-throughput biological datasets
The 2023 Annual General Meeting of the RSS Manchester local group will precede the seminar.
Abstract
Gaussian process (GP) models provide a popular approach for tackling a range of contemporary statistical problems in biology. They are attractive because they allow flexible modelling of temporal and spatial variation and they are easy to extend or combine into more complex models due to tractability of inference under linear transformations and marginalisation. I will present recent examples where we have applied hierarchical GP models [1], branching GP models [2] and differential equation GP models [3]. Applications include modelling protein melting curves [1], inferring single-cell pseudotime trajectories [2] and inferring mRNA degradation rates from time series data [3]. Our hierarchical and differential equation models have been re-implemented using newer differentiable programming frameworks (GPyTorch and GPflow) that allow better scalability and flexibility, enabling improved testing and inference schemes compared to earlier work.
[1] Le Sueur, C., Rattray, M., & Savitski, M. (2023). Hierarchical Gaussian process models explore the dark meltome of thermal proteome profiling experiments.
bioRxiv, 2023-10.
[2] Sarkans, E., Ahmed, S., Rattray, M., & Boukouvalas, A. (2022). Modelling sequential branching dynamics with a multivariate branching Gaussian process. Transactions on Machine Learning Research.
[3] Forbes Beadle, L., Love, J. C., Shapovalova, Y., Artemev, A., Rattray, M., & Ashe, H. L. (2023). Combined modelling of mRNA decay dynamics and single-molecule imaging in the Drosophila embryo uncovers a role for P-bodies in 5′ to 3′ degradation.
PLoS Biology,
21(1), e3001956.
Magnus Rattray is Professor of Computational & Systems Biology, and Director of the Institute for Data Science & Artificial Intelligence at the University of Manchester. He uses probabilistic modelling and Bayesian inference techniques to study biological systems across a broad range of temporal and spatial scales, from gene expression in single cells to longitudinal population health data. Recent work includes methods to uncover oscillations from single-cell imaging time course data and the development of scalable Gaussian process models for pseudotime and branching process inference using single-cell omics data.