Validated differentiation of Listeria monocytogenes serogroups by FTIR spectroscopy using an Artificial Neural Network based classifier in an accredited official food control laboratory

Abstract

Listeria monocytogenes is a well-known human pathogen, and especially the young, the elderly, otherwise immunocompromised individuals or pregnant women might suffer severe health consequences from listeriosis. Up to date, Fourier-Transform Infrared (FTIR) spectroscopical methods have been established for decades as a valuable means to differentiate between microbiological specimens at different taxonomical levels. In recent years, machine-based learning methods using Artificial Neural Networks (ANN) have highly advanced the discriminatory power of distinguishing spectrally closely related units such as serogroups of a given species. The present report describes the classification performance evaluation of a manufacturer (Bruker Daltonics, Bremen, Germany) - provided L. monocytogenes serogroup classifier by means of a formalized external validation carried out in a single laboratory. N = 630 absorption spectra from n = 94 food L. monocytogenes isolates pertaining to n = 11 serotypes / n = 3 serogroups were recorded on the IR Biotyper (Bruker Daltonics) and subsequently typed by the given classifier. The quantitative evaluation of inclusivity and exclusivity was performed following the principles of the Guidelines for Validating Species Identifications Using MALDI-TOF-MS issued by the German Federal Office of Consumer Protection (BVL) for a targeted identification. The FTIR classifier allocated all n = 486 spectra from n = 71 serogroup 1/2 and 4 isolates correctly to their respective serogroups, resulting in a true-positive rate of 100%. All remaining n = 144 spectra from n = 23 isolates of serogroup 3 were correctly allocated to an arbitrarily combined class entity of serogroups 3 and 7, likewise yielding both inclusivity and exclusivity rates of 100%. Consequently, in our official food control laboratory, this validated IR Biotyper method has been integrated into the accredited workflow for L. monocytogenes analysis in food samples according to ISO 11290, followed by MALDI-TOF MS confirmation on the species level to subsequent serogrouping and pre-selection by FTIR spectroscopy for Whole Genome Sequencing (WGS). This study confirmed that FTIR spectroscopy in combination with Artificial Neural Networks proves to be a reliable and thus valuable tool for the differentiation of the most common serogroups from Listeria monocytogenes. The application of FTIR spectroscopy saves valuable resources with respect to labor and time and thus facilitates outbreak analyses of the clinically relevant severe food-borne disease listeriosis where potentially a high number of isolates are involved.