Friday, 31 January 2020 | 12:00 noon
Professor Computational Bioinformatics, School of Informatics & SynthSys, Edinburgh, UK
Probabilistic machine learning applied to single-cell ‘omics: splicing analysis and epigenome clustering in single cells
(Host: S. Zacchigna)
Single-cell omics technologies promise to revolutionise our understanding of biology, yet they present formidable statistical challenges, due to the volume and sparsity of the data. In this talk, I will describe two recent strands of work involving the development of machine learning models tailored to single-cell omics data. In the first part of the talk, I will describe BRIE, Bayesian Regression for Isoform Estimation, a Bayesian approach to infer splicing ratios from single- cell RNA-seq data. BRIE works by learning a regression model that links sequence feature to inclusion ratios, thereby providing a statistically sound to partially pool data across genes and cells. In the second part, we will switch gear and discuss single-cell bisulfite sequencing data. I will present Melissa, a hierarchical Bayesian model to perform clustering and imputation of epigenomic data.
Refs: Huang and Sanguinetti, https://genomebiology.biomedcentral.com/articles/10.1186/s13059-017-1248-5 Kapourani and Sanguinetti, https://www.biorxiv.org/content/early/2018/05/02/312025
Short bio: Guido Sanguinetti is Chair of Computational Bioinformatics in the School of Informatics at the University of Edinburgh. His interests focus on machine learning methodologies and applications in biology, particularly in dynamical systems and high-throughput data modelling. He has published over 80 papers in international journals including Science, PNAS and Nature journals. He held an ERC Starting Grant 2012-2017 and was the recipient of the 2012 PNAS Cozzarelli Prize in Applied Science and Engineering. From 01/03/2020, he will take up a new role as Professor of Applied Physics and Chair of Data Science at SISSA.