Salmonella enterica and Escherichia coli are bacterial species that colonize different animal hosts with sub-types that can cause life-threatening infections in humans. Source attribution of zoonoses is an important goal for infection control as is identification of isolates in reservoir hosts that represent a threat to human health. In this study, host specificity and zoonotic potential were predicted using machine learning in which Support Vector Machine (SVM) classifiers were built based on predicted proteins from whole genome sequences. Analysis of over 1000 S. enterica genomes allowed the correct prediction (67 –90 % accuracy) of the source host for S. Typhimurium isolates and the same classifier could then differentiate the source host for alternative serovars such as S. Dublin. A key finding from both phylogeny and SVM methods was that the majority of isolates were assigned to host-specific sub-clusters and had high host-specific SVM scores. Moreover, only a minor subset of isolates had high probability scores for multiple hosts, indicating generalists with genetic content that may facilitate transition between hosts. The same approach correctly identified human versus bovine E. coli isolates (83 % accuracy) and the potential of the classifier to predict a zoonotic threat was demonstrated using E. coli O157. This research indicates marked host restriction for both S. enterica and E. coli, with only limited isolate subsets exhibiting host promiscuity by gene content. Machine learning can be successfully applied to interrogate source attribution of bacterial isolates and has the capacity to predict zoonotic potential.

Sequence analyses of pathogen genomes facilitate the tracking of disease outbreaks and allow relationships between strains to be reconstructed and virulence factors to be identified. However, these methods are generally used after an outbreak has happened. Here, we show that support vector machine analysis of bovine E. coli O157 isolate sequences can be applied to predict their zoonotic potential, identifying cattle strains more likely to be a serious threat to human health. Notably, only a minor subset (less than 10%) of bovine E. coli O157 isolates analyzed in our datasets were predicted to have the potential to cause human disease; this is despite the fact that the majority are within previously defined pathogenic lineages I or I/II and encode key virulence factors. The predictive capacity was retained when tested across datasets. The major differences between human and bovine E. coli O157 isolates were due to the relative abundances of hundreds of predicted prophage proteins. This finding has profound implications for public health management of disease because interventions in cattle, such a vaccination, can be targeted at herds carrying strains of high zoonotic potential. Machine-learning approaches should be applied broadly to further our understanding of pathogen biology. machine learning | zoonosis | Shiga toxin | E. coli | cattle

With the ever-expanding number of available sequences from bacterial genomes, and the expectation that this data type will be the primary one generated from both diagnostic and research laboratories for the foreseeable future, then there is both an opportunity and a need to evaluate how effectively computational approaches can be used within bacterial genomics to predict and understand complex phenotypes, such as pathogenic potential and host source. This article applied various quantitative methods such as diversity indexes, pangenome-wide association studies (GWAS) and dimensionality reduction techniques to better understand the data and then compared how well unsupervised and supervised machine learning (ML) methods could predict the source host of the isolates. The study uses the example of the pangenomes of 1203 Salmonella enterica serovar Typhimurium isolates in order to predict 'host of isolation' using these different methods. The article is aimed as a review of recent applications of ML in infection biology, but also, by working through this specific dataset, it allows discussion of the advantages and drawbacks of the different techniques. As with all such sub-population studies, the biological relevance will be dependent on the quality and diversity of the input data. Given this major caveat, we show that supervised ML has the potential to add real value to interpretation of bacterial genomic data, as it can provide probabilistic outcomes for important phenotypes, something that is very difficult to achieve with the other methods.