Metabolomics uses a high-throughput technique to analyze various metabolites and small molecules in biological samples to provide important insights into human health and disease. Untargeted metabolomics is one application that enables a comprehensive analysis of the metabolome, identifying crucial metabolites that indicate or contribute to health conditions. The advent of artificial intelligence (AI) and machine learning (ML) techniques has greatly improved untargeted metabolomics workflows, particularly in the field of high-resolution mass spectrometry exosomes which detects endogenous metabolites and exogenous chemicals in human tissue, correlating environmental exposures with disease outcomes.
Metabolism involves the body’s process of creating essential metabolites, encompassing catabolism (molecule breakdown for energy) and anabolism (compound synthesis required by cells). Metabolomics captures these endogenous metabolites and signaling molecules involved in gene expression, protein functions, and enzyme activity. A broader, semi-quantitative analysis of thousands of small molecules is provided by untargeted metabolomics. It’s part of a holistic approach called exposomics, which includes environmental exposures, diet, lifestyle, and psychosocial factors to reveal their impact on health. AI and ML are contributing to the detection and analysis of these complex data sets, improving the current understanding of chemical exposure and its impact on health.
For analyzing biological matrices, untargeted metabolomics uses techniques like LC or GC column chromatography, followed by HRMS detection and then data processing for the final analysis. Pre-and post-processing, data acquisition, and chemical identification are part of the workflow. AI and ML tools play a significant role in data processing, feature selection, and chemical identification to enrich the analysis of metabolomics data and enhance its biological interpretation.
AI and ML methods address the limitations of traditional models in defining the intricate structure of metabolomics data. AI and ML directly build and test models on data, revealing relationships between phenotypes, exposures, and diseases. AI and ML have been used in detecting diseases such as NAFLD, COVID-19, Alzheimer’s, and depression, proving their potential in metabolomics research.
AI and ML also assist in biomarker discovery, in which metabolite identification is crucial. It requires the annotation of chosen peaks using databases such as GNPS, Metlin, and the Human Metabolome Database. Spectral matching rates for specialized chemicals could be further improved, and advances in cognitive metabolomics and in silico tools are ameliorating identification accuracy.
Significant developments in untargeted chemical analysis and AI/ML tools have significantly reduced costs, enabling large-scale studies. They assist in data extraction, mining, and annotation, which are all essential in biomarker discovery. The main challenge is annotating unknown metabolites that are critical for biological interpretation. Current efforts are geared towards developing experimental databases and AI/ML models to enhance metabolite identification. There is a need to integrate biology-driven approaches with measurement-based methods to uncover obscure chemicals that impact health and catalyze discoveries in exosomes and precision health.
