NIST logo

Publication Citation: Improved Normalization of Systematic Biases Affecting Ion Current Measurements in Label-free Proteomics Data

NIST Authors in Bold

Author(s): Paul A. Rudnick; Xinjian Yan; Stephen E. Stein; Xia Wang; Nell Sedransk;
Title: Improved Normalization of Systematic Biases Affecting Ion Current Measurements in Label-free Proteomics Data
Published: February 21, 2014
Abstract: Systematic biases, or batch effects, in ,high-throughput‰ biological datasets can be described as obscuring variabilities introduced during experimental design or analytical processing. These biases, if left unresolved, can lead to misinterpretation of results. Batch effects in microarray experiments are well documented, and statistical methods for their removal (e.g., normalization) have been developed. But, in general, much less attention has been given to this topic in MS-based proteomics, particularly in understanding their underlying causes. ,Label-free‰ proteomic data present a particularly challenging case since signals from separate LC-MS runs are compared. This differs from labeling experiments in which 2 or more differently labeled samples are mixed and analyzed in a single LC run. In this work, we examined the multi-laboratory datasets from the first phase of NCI‰s CPTAC program, in which systematic biases were identified by monitoring signals arising from individual peptide ions. Surprisingly, the largest biases within labs were retention time and charge state, not intensity as is typically used during normalization. These effects were exaggerated in samples of unequal concentrations or spike-in levels. Our analysis indicates that the average precursor charge for peptides with higher charge state potentials is lower at higher sample concentrations. These effects are consistent with reduced relative protonation during ESI and demonstrate that the physical properties of the peptides themselves can serve as good reporters of systematic biases. Between labs, intensity was most commonly the ,top-ranked‰ bias variable, over retention time, precursor m/z and . A larger set of variables was then used to develop a stepwise normalization procedure. This statistical model was found to perform as well or better on the CPTAC mock biomarker data than other commonly used methods, and it does not require a priori knowledge of the systematic biases in a give
Citation: Molecular and Cellular Proteomics
Keywords: proteomics; mass spectrometry; bioinformatics; data normalization
Research Areas: Spectroscopy, Life Sciences Research, Mass
DOI: http://dx.doi.org/10.1074/mcp.M113.030593  (Note: May link to a non-U.S. Government webpage)