Listeria monocytogenes belongs to the category of facultative anaerobic bacteria, and is the pathogen of listeriosis, potentially lethal disease for humans. There are many similarities between L. monocytogenes and other non-pathogenic Listeria species, which causes great difficulties for their correct identification. The level of L. monocytogenes contamination in food remains high according to statistics from the Food and Drug Administration. This situation leads to food recall and destruction, which has caused huge economic losses to the food industry. Therefore, the identification of Listeria species is very important for clinical treatment and food safety. This work aims to explore an efficient classification algorithm which could easily and reliably distinguish Listeria species. We attempted to classify Listeria species by incorporating denoising autoencoder (DAE) and machine learning algorithms in matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF MS). In addition, convolutional neural networks were used to map the high dimensional original mass spectrometry data to low dimensional core features. By analyzing MALDI-TOF MS data via incorporating DAE and support vector machine (SVM), the identification accuracy of Listeria species was 100%. The proposed classification algorithm is fast (range of seconds), easy to handle, and, more importantly, this method also allows for extending the identification scope of bacteria. The DAE model used in our research is an effective tool for the extraction of MALDI-TOF mass spectrometry features. Despite the fact that the MALDI-TOF MS dataset examined in our research had high dimensionality, the DAE + SVM algorithm was still able to exploit the hidden information embedded in the original MALDI-TOF mass spectra. The experimental results in our work demonstrated that MALDI-TOF mass spectrum combined with DAE + SVM could easily and reliably distinguish Listeria species.
Keywords: Convolutional neural networks; Denoising autoencoder; Listeria species; MALDI-TOF; Machine learning; Support vector machines.
Copyright © 2021. Published by Elsevier B.V.