FSRN Project Summary: Using deep learning in protein interactions to detect a group of emerging bacterial pathogens
The Food Safety Research Network (FSRN) funded a partnering project to SRUC, to collaborate with STFC. The project was run Oct-2023 to Apr-2024. The slides present the summaries of the project.
The project examined whether computational approaches could be used to detect a group of bacteria with high risk of being pathogenic. Shiga-toxin producing Escherichia coli (STEC) are very diverse, with only some isolates able to cause serious human disease. They are mainly foodborne and occur in a range of hosts and habitats. Although characterised by carriage of Shiga toxin, there is a major challenge in determining which are likely to cause disease because multiple genetic factors are required. The current emergence of isolates with extensive genetic diversity has rendered the existing risk profiling tool as no longer fit-for-purpose.
Variation in STEC isolates occurs in both the total genetic complement and within single genes, which may indicate functional variation in the gene products. Protein variants with altered thermodynamic properties could be rendered sub-optimal for host interactions, effectively attenuated in their functions with reduced pathogenic risk potential. Our project aimed to determine whether protein functional variation could be exploited as a means to distinguish groups of isolates based on corresponding genetic differences. We used a set of 281 STEC isolates genomes (collected in a previous project), comprising clinical, i.e. associated with disease, and non-clinical isolates from wild deer, and food products (cheese, minced beef).
In Part 1, we used a well characterised STEC interacting protein pair, Intimin (Eae) and the translocated intimin receptor (Tir), as a proof of principle. In Part 2, we took a whole genome approach identifying virulence factors that are potentially associated with human disease, so increase the higher risk profile. We complied the findings into a flowchart to potentially classify pathogenic non-O157 STEC, based on PCR detection. Together this builds a far more nuanced and informative risk profile of any specific STEC isolate, using molecular approaches with a relatively rapid turn-around time, high throughput and with precedent for remote, non-lab based use.
Funding
UK Food Safety Research Network
Biotechnology and Biological Sciences Research Council
Find out more...