Batch profiling calibration for robust NMR metabonomic data analysis

Abstract

Metabonomic studies involve the analysis of large numbers of samples to identify significant changes in the metabolic fingerprints of biological systems, possibly with sufficient statistical power for analysis. While procedures related to sample preparation and spectral data acquisition generally include the use of independent sample batches, these might be sources of systematic variation whose effects should be removed to focus on phenotyping the relevant biological variability. In this work, we describe a grouped-batch profile (GBP) calibration strategy to adjust nuclear magnetic resonance (NMR) metabolomic data-sets for batch effects either introduced during NMR experiments or samples work-up. We show how this method can be applied to data calibration in the context of a large-scale NMR epidemiological study where quality control samples are available. We also illustrate the efficiency of a batch profile correction for NMR metabonomic investigation of cell extracts, where GBP can significantly improve the predictive power of multivariate statistical models for discriminant analysis of the cell infection status. The method is applicable to a broad range of NMR metabolomic/metabonomic cohort studies.