Statistical Analysis of Relative Labeled Mass Spectrometry Data from Complex Samples Using ANOVA

Abstract

Statistical tools enable unified analysis of data from multiple global proteomic experiments, producing unbiased estimates of normalization terms despite the missing data problem inherent in these studies. The modeling approach, implementation and useful visualization tools are demonstrated via case study of complex biological samples assessed using the iTRAQ™ relative labeling protocol. KeywordsProteomics; ANOVA; iTRAQ™; Normalization; relative labeling protocol; Missing data; GaussSiedel; Backfitting; Fixed effects model; Mixed effects model A. INTRODUCTIONThe objective of global proteomics via mass spectrometry is to detect and quantify all proteins present in a biological sample. Proteins that exhibit an increase/decrease in abundance between two or more groups of interest, (e.g., diseased and non-diseased) are considered candidate CORRESPONDING AUTHOR FOOTNOTE Ann L. Oberg, Mayo Clinic, Cancer Center Statistics, 200 First St SW, Rochester, MN 55905. Telephone (507) 538-1556; Fax (507) biomarkers. However, experimental factors such as differences in sample collection, sample characteristics such as cellular concentration, variations in sample processing and the experimental process add variability to the observed abundances. Experimental variability hinders the comparison of effects of interest, and if not accounted for during the design and analysis stages, can lead the researcher down an erroneous path of discovery.Several mass spectrometry (MS) techniques have been developed that allow greater control over experimental factors that introduce variability and ultimately decrease the quality of the data. Recently, focus has centered on the ability to assess multiple samples within a single MS experiment. Binary sample labeling techniques such as 16 O/ 18 O (1) , ICAT™ (2) , and SILAC (3) were developed to evaluate paired samples whereas iTRAQ™ (4) was developed to simultaneously analyze four, and more recently, eight samples (5) . The binary labeling techniques add complexity to the acquired spectra and to their interpretation by introducing additional peaks into the mass spectra. Furthermore, overlapping isotopic clusters require further analytical techniques to deconvolute the resulting spectrum and the associated protein/ peptide abundances (4,6,7) . The iTRAQ™ labeling system overcomes this to some extent since the labeled species are isobaric and protein abundances are measured only in the resulting MS/ MS fragmentation spectra.Although sample labeling techniques allow greater control over experimental variability within an MS experiment, the analysis of multiple MS experiments remains difficult. Within an experiment, it is important that equal amounts of total protein are labeled under each labeling condition to ensure that the observed abundances are not influenced by total protein concentration. Once the samples are labeled and mixed together for MS analysis, labeling methods naturally control for instrument variability. The same principles apply when performing multiple experiments; with th...

Huntingtin protein interactions altered by polyglutamine expansion as determined by quantitative proteomic analysis

“…27 To identify proteins most significantly changed in abundance between normal and expanded Htt complexes, we used statistical analysis of variance (ANOVA) approach, which considers both biologic and experimental sources of variability. Relative peptide and protein abundances were estimated based on collected reporter ion peak areas from all observed tandem mass spectra using linear mixed effects models [34][35][36] (and Herbrich et al in preparation). This allows to correct for different amount of material loaded in the channels (as a random effect) and enables a comparison between the relative protein abundances (as random effects), comparing proteins present within expanded and normal Htt complexes.…”

“…Relative peptide and protein abundances were estimated based on collected reporter ion peak areas from all observed tandem mass spectra using linear mixed effects models 34,35 (and Herbrich et al manuscript in preparation). Multiple comparisons were addressed by controlling the family-wise error rate via Bonferroni correction.…”

A Statistical Model for iTRAQ Data Analysis

“…For example, Jain et al 18 state that extremely large outlying expression values can result in poorly powered tests of differential expression, while chance occurrences of genes with similar expression patterns across replicates can lead to increased false positive rates. To mitigate these potential problems 10 , they propose the use of a locally pooled error estimate for their gene-specific differential expression analyses, while Tusher et al 19 add a constant to all error estimates in their popular 'Statistical Analysis of Microarrays' method. For proteomics applications, analysis of proteins one at a time may alleviate some of the computational burden, in particular when combining data across iTRAQ experiments.…”

“…The GPS software calculates the average ratio, GM i , for protein i as (9) where j = 1,...,J i and GM i is the geometric mean of the peptide ratios. The standard deviation of the J i peptide ratios, SD i , for protein i is calculated by the GPS software as (10) where sd i is the standard deviation of the log peptide ratios, {log(X j )}.…”

“…Specifically, a denominator is selected by the user from among the four tags, peptide ratios subsequently are constructed for each tag pair (three in all), and the corresponding protein ratios and standard deviations are calculated from these peptide ratios based on Equations (9) and (10). Typically, GPS software normalizes peptide ratios prior to protein quantitation based on the assumption that the majority of ratios are equal to one.…”