Novel protocol for metabolomics data normalization and biomarker discovery in human tears.

Abstract

Objectives: Human tear analysis holds promise for biomarker discovery, but its clinical utility is hindered by the lack of standardized reference values, limiting interindividual comparisons. This study aimed at developing a protocol for normalizing metabolomic data from human tears, enhancing its potential for biomarker identification.<strong> Methods: </strong> Tear metabolomic profiling was conducted on 103 donors (64 females, 39 males, aged 18-82 years) without ocular pathology, using the AbsoluteIDQ™ p180 Kit for targeted metabolomics. A predictive normalization model incorporating age, sex, and fasting time was developed to correct for interindividual variability. Key metabolites from six compound families (amino acids, biogenic amines, acylcarnitines, lysophosphatidylcholines, phosphatidylcholines, and sphingomyelins) were identified as normalization references. The approach was validated using Linear Discriminant Analysis (LDA) to test its ability to classify donor sex based on metabolite concentrations.<strong> Results: </strong> Metabolite concentrations exhibited significant interindividual variability. The normalization model, which predicted metabolite concentrations based on a reference "concomitant" metabolite from each compound family, successfully reduced this variability. Using the ratio of observed-to-predicted concentrations, the model enabled robust comparisons across individuals. LDA classification of donor sex using acylcarnitine C4 achieved 78 % accuracy, correctly identifying 92 % of female donors. This approach outperformed traditional statistical and machine learning methods (Lasso logistic regression and Random Forest classification) in sex discrimination based on tear metabolomics.<strong> Conclusions: </strong> This novel normalization protocol significantly improves the reliability of tear metabolomics by enabling standardized interindividual comparisons. The approach facilitates biomarker discovery by mitigating variability in metabolite concentrations and may be extended to other biological fluids, enhancing its applicability in precision medicine.