Normalization strategies in neonatal steroid metabolomics: a comparative analysis of probabilistic quotient and peer group approaches.

Abstract

Introduction: Steroid metabolomics in neonatal populations is challenged by considerable physiological heterogeneity and technical variability, which complicate the interpretation and comparability of metabolite profiles. Effective normalization strategies are essential to ensure accurate data analysis in this context.

Material and methods: We analyzed 24-hour urinary steroid profiles in a cohort of 50 neonates (including very preterm, late preterm, and full-term infants) using gas chromatography-mass spectrometry. Two normalization techniques were compared: probabilistic quotient normalization (PQN) and peer group normalization (PGN). Normalization performance was assessed via distribution metrics, correlation with anthropometric variables, and principal component analysis (PCA).

Results: PGN achieved superior distributional normalization, with 27 of 30 metabolites conforming to normality assumptions, compared to 21 using PQN. PGN also eliminated all significant correlations between steroid levels and anthropometric parameters, indicating effective reduction of physiological confounding. In contrast, PQN partially mitigated such associations but was less robust in handling high-abundance metabolites. PCA confirmed improved sample dispersion and group separation after normalization, with method-dependent differences in Scores Plot.

Conclusions: Peer group normalization is a sophisticated approach to reducing physiological variability in neonatal steroid profiling. These observations lend further credence to PGN as a promising strategy for standardizing steroid metabolomics in the field of neonatology. Nevertheless, further validation is necessary to substantiate these findings.