Comprehensive assessment of the diminished statistical power caused by nonlinear electrospray ionization responses in mass spectrometry-based metabolomics

Abstract

It seems to be well received that nonlinear electrospray ionization (ESI) distorts the signal distribution in mass spectrometry (MS) analysis, thus leading to diminished statistical power for t-test. However, the exact consequence and possible solutions to this quantitative issue have not been systematically explored. In this work, using a serial diluted urine metabolomics dataset, we demonstrated that over 80% of the metabolic features present nonlinear ESI response patterns, causing either left-skewed or right-skewed MS signal distributions. Among them, right-skewed MS distributions cannot be rescued using conventional data transformation (e.g., log transformation, power transformation). Furthermore, using a Monte Carlo simulation, we quantitatively assessed the reduced statistical power for t-test calculated using MS signal data in various sample sizes and effect sizes. In all these comparisons, t-test using MS signal data has consistently lower statistical power than t-test using metabolic concentration data. To address this statistical issue, we proposed a bioinformatic workflow, termed PowerU, to minimize the diminished statistical power caused by both the nonlinear ESI response and the intrinsic non-normal distribution of metabolic concentrations. The PowerU workflow is composed of two steps. The first step is to convert MS signals to quality control (QC) sample injection amounts to solve the skewed MS signal distributions. The second step is to perform a Shapiro-Wilk test to determine data normality and then use the normality results to guide the application of t-test and Mann-Whitney U test for the best statistical outcome. PowerU was tested using a metabolomics study of mouse cecum samples. Results demonstrate that the PowerU workflow can significantly boost statistical power for t-test and facilitate the discovery of significantly altered metabolites for downstream biological interpretation.