Rapid detection of viable microbes remains a challenge in fields such as microbial food safety. We here present the application of deep learning algorithms to the rapid detection of pathogenic and non-pathogenic microbes using metabolomics data. Microbes were incubated for 4 h in a protein-free defined medium, followed by 1D 1H nuclear magnetic resonance (NMR) spectroscopy measurements. NMR spectra were analyzed by spectral binning in an untargeted metabolomics approach. We trained multilayer (“deep”) artificial neural networks (ANN) on the data and used the resulting models to predict spectra of unknown microbes. ANN predicted unknown microbes in this laboratory setting with an average accuracy of 99.2% when using a simple feature selection method. We also describe learning behavior of the employed ANN and the optimization strategies that worked well with these networks for our datasets. Performance was compared to other current data analysis methods, and ANN consistently scored higher than random forest models and support vector machines, highlighting the potential of deep learning in metabolomics data analysis.

Aims: To develop a broadly applicable medium free of proteins with welldefined and reproducible chemical composition for the cultivation of various micro-organisms with food safety significance. Methods and Results: The defined medium was designed as a buffered minimal salt medium supplemented with amino acids, vitamins, trace metals and other nutrients. Various strains commonly used for food safety research were selected to test the new defined medium. We investigated single growth factors needed by different strains and the growth performance of each strain cultivated in the defined medium. Results showed that the tested strains initially grew slower in the defined medium compared to tryptic soy broth, but after an overnight incubation cultures from the defined medium reached adequately high cell densities. Conclusions: The newly designed defined medium can be widely applied in food safety studies that require media with well-defined chemical constituents. Significance and Impact of the Study: Defined media are important in studies of microbial metabolites and physiological properties. A defined medium capable of cultivating different strains simultaneously is needed in the food safety area. The new defined medium has broader applications in comparing different strains directly and provides more reproducible results.

Introduction Artificial Neural Networks (ANN) are increasingly used in metabolomics. Objectives Given the multitude of implementations of ANN, there is no straightforward way to identify important features (metabolites). We developed a simple numeric score, the FIA score, to identify features of high importance. Methods FIA analysis was implemented in R and tested on microbial and human datasets. Results FIA scores correlated significantly to p -values and can provide information on the stability of ANN models. Conclusion FIA scores are a novel, simple score to assess the impact of features that will help interpreting ANN outcomes in the metabolomics area.

No abstract