Deep Learning Models for LC-MS Untargeted Metabolomics Data Analysis

Metabolomics, the measurement of all metabolites in a given system, is a growing research field with great potential and manifold applications in precision medicine. However, the high dimensionality and complexity of metabolomics data requires expert knowledge, the use of proper methodology, and is largely based on manual interpretation. In this book chapter, we discuss recent published approaches using deep learning to analyze untargeted metabolomics data. These approaches were applied within diverse stages of metabolomics data analysis, e.g. to improve preprocessing, feature identification, classification, and other tasks. We focus our attention on deep learning methods applied to liquid chromatography mass spectrometry (LC-MS), but these models can be extended or adjusted to other applications. We highlight current deep learning-based computational workflows that are paving the way toward high(er)-throughput use of untargeted metabolomics, making it effective for clinical, environmental and other types of applications.