Hosted by the Interactive AI Centre for Doctoral Training and the Intelligent Systems Lab
Contact iai-cdt@bristol.ac.uk if you would like lunch, served between 13.30-14.00
Abstract: The identification of epitopes, which are the specific portions of proteins recognised as 'foreign' by our immune system, is an important step in the development of diagnostic tests, vaccines and immunotherapies. Currently, most machine learning approaches for epitope prediction are developed under a generalist, one-size-fits-all approach - meaning that they are trained on data from a wide range of pathogens and are intended to generate predictions for proteins coming from any pathogen. Although successful to some extent, this approach results in the need for very large models and can lead to poor generalisation when deployed to generate predictions for emerging or under-studied pathogens. In this talk, I'll present recent work on the development and validation of phylogeny-aware machine learning models optimised for specific (groups of) pathogens. This is made possible by leveraging available data from evolutionarily related organisms, using a taxonomic scaffold. I'll present current results, including a case study of identifying new diagnostic targets for monkeypox and related viruses, and discuss some promising directions of research.