Rapid and accurate detection of Shiga toxin-producing Escherichia coli (STEC) serotype O157 : H7 by mass spectrometry directly from the isolate, using 10 potential biomarker peaks and machine learning predictive models.

The different pathotypes of can produce a large number of human diseases. Surveillance is complex since their differentiation is not easy. In particular, the detection of Shiga toxin-producing (STEC) serotype O157 : H7 consists of stool culture of a diarrhoeal sample on enriched and/or selective media and identification of presumptive colonies and confirmation, which require a certain level of training and are time-consuming and expensive. Matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) is a quick and easy way to obtain the protein spectrum of a microorganism, identify the genus and species, and detect potential biomarker peaks of certain characteristics. To verify the usefulness of MALDI-TOF MS to rapidly identify and differentiate STEC O157 : H7 from other pathotypes. The direct method was employed, and the information obtained using Microflex LT platform-based analysis from 60 clinical isolates (training set) was used to detect differences between the peptide fingerprints of STEC O157 : H7 and other strains. The protein profiles detected laid the foundations for the development and evaluation of machine learning predictive models in this study. The detection of potential biomarkers in combination with machine learning predictive models in a new set of 142 samples, called 'test set', achieved 99.3 % (141/142) correct classification, allowing us to distinguish between the isolates of STEC O157 : H7 and the other group. Great similarity was also observed with respect to this last group and the species when applying the potential biomarkers algorithm, allowing differentiation from STEC O157 : H7 Given that STEC O157 : H7 is the main causal agent of haemolytic uremic syndrome, and based on the performance values obtained in the present study (sensitivity=98.5 % and specificity=100.0 %), the implementation of this technique provides a proof of principle for MALDI-TOF MS and machine learning to identify biomarkers to rapidly screen or confirm STEC O157 : H7 versus other diarrhoeagenic in the future.