Abstract / Summary
The introduction of metabolomics into depression research, together with advances in bioinformatics, has substantially expanded our understanding of molecular alterations associated with this disorder. However, the rapid growth of metabolomic studies, combined with variability in analytical workflows, biological matrices, and experimental models, presents major challenges for data interpretation and cross-study integration. In this article, we systematically summarized reported metabolite alterations in human and animal studies using a combined natural language processing (NLP)-based text-mining and vote-counting approach. To our knowledge, this is the first study to apply NLP-based abstract mining to systematically analyze depression-related metabolomics literature across multiple biological compartments. Text-mining analysis of publications retrieved from PubMed and Scopus identified proteinogenic amino acids as the most consistently reported class of altered metabolites across peripheral and central matrices, followed by lipids and other metabolic categories. Vote-counting analysis of non-targeted metabolomics studies revealed relatively consistent decreases in circulating large neutral and branched-chain amino acids, with Leu, Met, Trp, Tyr, and Val representing the most consistently overlapping alterations across human and animal studies. Finally, these findings were interpreted in the context of brain barrier transport systems, highlighting how altered metabolite exchange across the blood-brain interfaces may contribute to compartment-specific metabolic alterations observed in depression.