Identifying Drug-Induced Metabolic Signatures in Comparative Untargeted NMR Metabolomics Using Difference Spectra

Untargeted NMR-based metabolomics is widely applied to investigate drug-induced metabolic alterations. However, comparative analyses across different biological models are often limited by dominant baseline variability arising from intrinsic metabolic differences and distinct culture conditions. This variability may obscure treatment-related signatures and compromise both multivariate and univariate statistical analyses. Here, we propose a data-processing strategy based on difference spectra as input for statistical modeling. For each treatment condition, difference spectra are first calculated by subtracting each paired control binned spectrum from its corresponding treated binned spectrum. These replicate-level binned difference spectra are then averaged to reduce intersample variability while enhancing the detection of treatment-induced metabolic changes within an untargeted framework. The strategy was validated by testing four structurally diverse gold(I) compounds in two ovarian cancer cell lines characterized by significant differences in basal metabolism and grown in distinct media. While conventional multivariate analysis of individual spectra failed to extract shared drug-related signatures, the difference-spectra approach enabled the identification of conserved metabolic perturbations. Altered pathways across cell models were then evaluated based on univariate analysis. Overall, the proposed method provides a robust and broadly applicable analytical tool for comparative untargeted metabolomics, facilitating the identification of common metabolic fingerprints despite substantial baseline heterogeneity.