Publication Abstract

Random Forest based source attribution models were developed from a ‘one health’ resource of 4,230 high-quality whole genome assemblies from E. coli isolated from a wide range of sources, predominantly originating in Scotland, including wastewater, human clinical, livestock, canine clinical, food and shellfish. The models allowed probabilistic assignment of E. coli isolates from food and water samples to potential sources of contamination.  The models identified samples likely contaminated with E. coli from human sources although isolates with sequence types (STs) strongly associated with human urinary tract and bacteraemia infections were only rarely found in food samples, <5% of the E. coli analysed.  However, livestock-human co-attribution of E. coli in food samples was common and consistent with routine human exposure to only specific subsets of livestock E. coli, potentially a result of selection in the food chain. Notably, ~16% of human UTI and bacteraemia isolates could be attributed to livestock sources using the models. Overall, this research shows the potential of source attribution models to provide insights into E. coli transmission through the food chain in a national setting.

Antonia Chalka^1,2, Louise Crozier², Adriana Vallejo-Trujillo^1,2, Alison Low¹, Sean McAteer¹, Vesa Qarkaxhija¹, Kate Templeton³, Sue C. Tongue^4a, Judith Evans^4a, Geoff Foster^4b, Catriona Webster^4a, Thomas Evans⁵, Charis Marwick⁶, Ahmed Raza⁶, Benjamin J. Parcell⁶, Matthew T. G. Holden⁶, Tom McNeilly⁷, Stephen Fitzgerald⁷, Mairi Mitchell⁷, Nuno Silva⁷, Emily Robertshaw-McFarlane⁷, Scott Hamilton⁷, Beth Wells⁷, Clare Hamilton⁷, Eleanor Watson⁷, David Findlay⁸, Peter Singleton⁸, Julie Bolland⁸, John Redshaw⁸, David Walker⁹, Jane Heywood⁹, Charlotte King⁹, Craig Baker-Austin⁹, Athina Papadopoulou⁹, Andy Powell⁹, Genever Morgan¹⁰, Gavin Paterson¹, Jacqui Mcelhiney², David L. Gally^1,2