Identification of differentially expressed peptides in high-throughput proteomics data

Abstract

With the advent of high-throughput proteomics, the type and amount of data pose a significant challenge to statistical approaches used to validate current quantitative analysis. Whereas many studies focus on the analysis at the protein level, the analysis of peptide-level data provides insight into changes at the sub-protein level, including splice variants, isoforms and a range of post-translational modifications. Statistical evaluation of liquid chromatography-mass spectrometry/mass spectrometry peptide-based label-free differential data is most commonly performed using a t-test or analysis of variance, often after the application of data imputation to reduce the number of missing values. In high-throughput proteomics, statistical analysis methods and imputation techniques are difficult to evaluate, given the lack of gold standard data sets. Here, we use experimental and resampled data to evaluate the performance of four statistical analysis methods and the added value of imputation, for different numbers of biological replicates. We find that three or four replicates are the minimum requirement for high-throughput data analysis and confident assignment of significant changes. Data imputation does increase sensitivity in some cases, but leads to a much higher actual false discovery rate. Additionally, we find that empirical Bayes method (limma) achieves the highest sensitivity, and we thus recommend its use for performing differential expression analysis at the peptide level.

Phosphoproteomic analysis reveals plant DNA damage signalling pathways with a functional role for histone H2AX phosphorylation in plant growth under genotoxic stress

“…Quantitative MS identified 2220 phosphopeptides and 4798 phosphorylation sites (Table S1). 225 phosphopeptides (180 proteins, 254 sites) were identified that displayed significant change in abundance upon X‐irradiation of wild‐type lines (Table S2), analyzing four independent replicates for each treatment and correcting for a false discoveries using limma in R (van Ooijen et al ., ). Of the 254 X‐ray responsive phosphorylation sites identified here, only six were previously identified in the DNA damage signalling network of mature wild‐type plants (Table S2), highlighting the wide range of phosphoproteins induced upon genotoxin treatment.…”

Proteomic Profiling of Mouse Helper T Cell Differentiation

“…Consequently, the proteins that are detected differ from run to run, resulting in missing quantifications in the (replicate values of the) data set. Although these can be imputed, we recently showed that this leads to high levels of false positives and hence, is not recommended for high‐throughput protein analysis . Future proteomics experiments will yield additional proteins involved in Th skewing, but the observation that only a small fraction of all (quantified) proteins are associated with Th differentiation is in line with our genome wide mRNA profiling experiments.…”

“…The mouse/spleen group from which the cells were derived was used as a co‐variate to control for replicate associated effects, for details see refs. and . The intensity data was processed using custom R scripts: after log 2 ‐transformation, the sample medians were shifted onto a single median value (Figure S3, Supporting Information).…”

CAncer bioMarker Prediction Pipeline (CAMPP) - A standardised and user-friendly framework for the analysis of quantitative biological data

“…For the purpose of variable selection CAMPP employs limma (linear models for microarray data) for differential expression/abundance analysis (DEA, DAA) (Ritchie, et al, 2015). Although limma was originally designed for analysis of microarray data and subsequently revised to handle RNA sequencing data, this software is very flexible and has recently been shown to perform very well with quantitative mass spectrometry data (Kammers, et al, 2015;van Ooijen, et al, 2018). In addition to being versatile, limma has been shown to work exceptionally well on datasets with small sample sizes (Seyednasrollah, et al, 2015;Soneson and Delorenzi, 2013).…”