Psychotropic drug-induced weight gain and untargeted metabolomics: machine learning-driven results from a prospective cohort study

Several psychotropic drugs can induce important weight gain. The present study aims to elucidate the metabolomic fingerprint of this adverse effect. From the Psymetab cohort, 151 follow-ups from 135 patients starting a weight-inducing psychotropic drug were selected. Untargeted metabolomics analyses were applied to plasma samples collected at treatment baseline and at another time point during the follow-up. Dimension reduction and feature selection approaches such as Minimum-Redundancy/Maximum-Relevance coupled with a bootstrap of 5000 Least Absolute Shrinkage and Selection Operator were applied to select metabolic features associated with weight change per month of treatment. Our selection algorithm identified 56 metabolic features. Out of the 56, 10 were identified, of which 6 were confirmed by standard references. Consistent with previous studies, increases in the intensity changes of glycerophospholipids such as lysophosphatidylcholine (18:2/0:0), and tryptophan metabolites such as acetylkynurenine and 5-hydroxy-L-tryptophan were positively associated with weight change. Moreover, increases in the intensity changes of the amino acids citrulline and N-Acetyl-L-aspartic, and of the xanthine 1-methyluric acid and of the pyrimidine nucleoside uridine were positively associated with weight changes, whereas increases in the intensity changes of the amino acid glutamine and of the steroidal glycoside glycochenodeoxycholic acid 3-glucuronide were negatively associated with the outcome. The present study highlighted novel potential biomarkers for weight change under psychotropic treatment, confirming an alteration in the pathways of glycerophospholipids and tryptophan metabolites. These results emphasise the potential use of metabolomics for clarifying metabolic changes and for defining biomarkers with a prospective use in clinical practice.