Development and Evaluation of Normalization Methods for Label-free Relative Quantification of Endogenous Peptides

The performances of 10 different normalization methods on data of endogenous brain peptides produced with label-free nano-LC-MS were evaluated. Data sets originating from three different species (mouse, rat, and Japanese quail), each consisting of 35–45 individual LC-MS analyses, were used in the study. Each sample set contained both technical and biological replicates, and the LC-MS analyses were performed in a randomized block fashion. Peptides in all three data sets were found to display LC-MS analysis order-dependent bias. Global normalization methods will only to some extent correct this type of bias. Only the novel normalization procedure RegrRun (linear regression followed by analysis order normalization) corrected for this type of bias. The RegrRun procedure performed the best of the normalization methods tested and decreased the median S.D. by 43% on average compared with raw data. This method also produced the smallest fraction of peptides with interblock differences while producing the largest fraction of differentially expressed peaks between treatment groups in all three data sets. Linear regression normalization (Regr) performed second best and decreased median S.D. by 38% on average compared with raw data. All other examined methods reduced median S.D. by 20–30% on average compared with raw data.Peptidomics is defined as the analysis of the peptide content within an organism, tissue, or cell (1–3). The proteome and peptidome have common features, but there are also prominent differences. Proteomics generally identifies proteins by using the information of biologically inactive peptides derived from tryptic digestion, whereas peptidomics tries to identify endogenous peptides using single peptide sequence information only (4). Endogenous neuropeptides are peptides used for intracellular signaling that can act as neurotransmitters or neuromodulators in the nervous system. These polypeptides of 3–100 amino acids can be abundantly produced in large neural populations or in trace levels from single neurons (5) and are often generated through the cleavage of precursor proteins. However, unwanted peptides can also be created through post-mortem induced proteolysis (6). The later aspect complicates the technical analysis of neuropeptides as post-mortem conditions increase the number of degradation peptides. The possibility to detect, identify, and quantify lowly expressed neuropeptides using label-free LC-MS techniques has improved with the development of new sample preparation techniques including rapid heating of the tissue, which prevents protein degradation and inhibition of post-mortem proteolytic activity (7, 8).It has been suggested by us (4, 5) and others (9) that comparing the peptidome between samples of e.g. diseased and normal tissue may lead to the discovery of biologically relevant peptides of certain pathological or pharmacological events. However, differences in relative peptide abundance measurements may not only originate from biological differences but also from systematic bias and noise. To reduce the effects of experimentally induced variability it is common to normalize the raw data. This is a concept well known in the area of genomics studies using gene expression microarrays (10–12). As a consequence, many methods developed for microarray data have also been adapted for normalizing peptide data produced with LC-MS techniques (10–16). Normally the underlying assumption for applying these techniques is that the total or mean/median peak abundances should be equal across different experiments, in this case between LC-MS analyses. Global normalization methods refer to cases where all peak abundances are used to determine a single normalization factor between experiments (13, 15, 16), a subset of peaks assumed to be similarly abundant between experiments (16) is used, or spiked-in peptides are used as internal standards. In a study by Callister et al. (14), normalization methods for tryptic LC-FTICR-MS peptide data were compared. The authors concluded that global or iterative linear regression works best in most cases but also recommended that the best procedure should be selected for each data set individually. Methods used for normalizing LC-MS data have been reviewed previously (14, 17, 18), but to our knowledge only Callister et al. (14) have used small data sets to systematically evaluate such methods. None of these studies have targeted data of endogenous peptides.In this study, the effects of 10 different normalization methods were evaluated on data produced by a nano-LC system coupled to an electrospray Q-TOF or linear trap quadrupole (LTQ)¹ mass spectrometer. Normalization methods that originally were developed for gene expression data were used, and one novel method, linear regression followed by analysis order normalization (RegrRun), is presented. The normalization methods were evaluated using three data sets of endogenous brain peptides originating from three different species (mouse, rat, and Japanese quail), each consisting of 35–45 individual LC-MS analyses. Each data set contained both technical and biological replicates.