UPLC-MS and multivariate analysis reveal metabolic pathway adaptations to training in professional football players.

Metabolomics provides direct insights into biological processes by analyzing metabolites. While univariate and multivariate analyses, alongside pathway and functional analysis tools like mummichog, are commonly employed, integrating these results to interpret biological significance remains a challenge, limiting the potential of metabolomic analyses. This study introduces innovative methods to analyze metabolic adaptations in professional football players using a unique UPLC-TOF-MS dataset comprising 93 urinary samples collected over a 10-month football season. Urinary metabolomic profiles were linked to training load data obtained through an electronic performance tracking system. Three approaches combining multivariate analysis with pathway-level insights were developed. PLS regression p-values integrated with functional metabolic analysis identified training load-associated pathways overlooked by univariate methods. Cluster cross-validation enhanced these insights by assessing the contribution of each pathway to the predictive performance, ranking pathways driving the PLS model. Backward feature elimination refined metabolic features most strongly linked to training load, improving the practicality of findings for targeted biomarker validation. Univariate analyses highlighted alterations in Phenylalanine and Histidine metabolisms related to total external load. Multivariate methods identified additional pathways, including Tryptophan, Purine, and Tyrosine metabolisms, as top contributors to the association between metabolic profiles and training load. Results demonstrate that combining multivariate techniques with functional analysis expands understanding of athletes' metabolic responses, offering more comprehensive biomarker discovery beyond the scope of univariate approaches. These findings underscore the value of integrating multivariate strategies with pathway insights to enhance the biological interpretation of metabolomic data.