Targeted Metabolomic Profiling of Emotional and Reflex Tears: A Paired Exploratory Study

Objectives: Emotional tears are generated in a distinct neurophysiological context from reflex tears, but their metabolic composition remains poorly understood. Methods: This exploratory paired targeted-metabolomics study compared emotional and reflex tears collected from 22 healthy volunteers using a 600-multiple-reaction-monitoring platform. Among 412 detected metabolites, 344 were retained after data preprocessing. Paired statistical analysis prioritized 23 candidate metabolites based on the combined criteria of unadjusted p-value, fold change, and consistency of within-subject change. Results: Seven candidates were lower, and 16 were higher in emotional tears. Salicylic acid was retained in the descriptive candidate set but excluded from the primary machine-learning analysis because a contribution from the reflex-tear induction procedure could not be ruled out. Among the remaining 22 candidates, 4-hydroxy-3-methylbenzoic acid, vanillic acid, cytidine-5′-monophosphate, and hydroxyphenyllactic acid were consistently ranked among the leading features. A fixed four-metabolite combination achieved an area under the receiver operating characteristic curve of 0.864 (95% CI, 0.756–0.957) under ordinary leave-one-subject-out cross-validation. When candidate screening, feature selection, and model fitting were repeated within each training fold, the best nested pipeline achieved an area under the curve of 0.725 (95% CI, 0.603–0.843). Pathway mapping further linked the candidate metabolites to histidine, tyrosine, fatty-acid, ether-lipid, and ubiquinone-related metabolism. Conclusions: These findings demonstrate measurable within-subject metabolic differences between emotional and reflex tears and identify a focused set of candidate metabolites for future validation. Because no individual metabolite remained significant after false discovery rate correction and no independent validation cohort was available, the candidate signals and discrimination models should be confirmed in larger, independently collected cohorts.