Joe Hilton (University of Manchester) and Liam Brierley (University of Glasgow)

Speaker: Joe Hilton (University of Manchester)
Title: Identifying environmental and ecological drivers of highly pathogenic avian influenza infections in European wild birds
Abstract:
In this study we expand upon existing SDM approaches to HPAI risk mapping by explicitly modelling the effect of ecological variables such as prevalence of HPAI host taxa, species richness, and diet preference. We use Bayesian additive regression trees, a non-parametric machine learning method, to model HPAI presence at a 10km resolution. We train our model using a Europe-wide dataset of over 8,000 geospatial records of HPAI from 53 wild bird families. We optimise our model using 5-fold cross-validation and validate on outbreaks excluded from model training.

Speaker: Liam Brierley (University of Glasgow)
Title: Inferring the next zoonotic spillover of avian influenza directly from genomic machine learning
Abstract:
Avian influenza (AI) remains a serious threat to both livestock and human health. Human cases in the current highly pathogenic H5N1 outbreak have been rare. However, concerns remain around potential for new lineages to emerge that spill over and transmit between humans more readily, particularly in the wake of potential transmission within cattle.
I use NCBI GenBank and GISAID to source >18,000 whole genome sequences of AI from 122 subtypes including 14 with known zoonotic events. To prevent over-fitting models to well-sampled lineages, sequences are collapsed into ~4,000 non-zoonotic clusters and ~100 zoonotic clusters based on % shared sequence identity before model training at cluster-level.