Evaluation of linear models and missing value imputation for the analysis of peptide-centric proteomics

Abstract

BackgroundSeveral methods to handle data generated from bottom-up proteomics via liquid chromatography-mass spectrometry, particularly for peptide-centric quantification dealing with post-translational modification (PTM) analysis like reversible cysteine oxidation are evaluated. The paper proposes a pipeline based on the R programming language to analyze PTMs from peptide-centric label-free quantitative proteomics data.ResultsOur methodology includes variance stabilization, normalization, and missing data imputation to account for the large dynamic range of PTM measurements. It also corrects biases from an enrichment protocol and reduces the random and systematic errors associated with label-free quantification. The performance of the methodology is tested by performing proteome-wide differential PTM quantitation using linear models analysis (limma). We objectively compare two imputation methods along with significance testing when using multiple-imputation for missing data.ConclusionIdentifying PTMs in large-scale datasets is a problem with distinct characteristics that require new methods for handling missing data imputation and differential proteome analysis. Linear models in combination with multiple-imputation could significantly outperform a t-test-based decision method.Electronic supplementary materialThe online version of this article (10.1186/s12859-019-2619-6) contains supplementary material, which is available to authorized users.

Proteome-Wide Analysis of Cysteine Reactivity during Effector-Triggered Immunity

“…The data analysis pipeline consists of: (1) data normalization and variance-stabilization, (2) data filtering, (3) missing data imputation, (4) linear model analysis (limma), (5) differential analysis using limma output and fold change criteria, (6) multiple imputation and binomial testing, (7) time-series clustering analysis, (8) structure-reactivity clustering analysis of Cys residues, and (9) functional Gene Ontology (GO) annotation analysis. The first peptide analysis steps follow the implementation of the peptide-centric proteomics pipeline described in Berg et al (2019), which has demonstrated very good performance for analysis of reversible oxidized cysteines on a benchmark dataset. An important result from our previous study was the significant performance improvement obtained when using a linear model in combination with multiple data imputations for comparative analysis of large-scale reversible Cys oxidation datasets.…”

“…Detailed evaluation of the steps used in the enrichment of reversible Cys oxidation was performed as described in Berg et al (2019). Briefly, each replicate and 50 mg of TPS6B (GE Healthcare Bio-Sciences) resin slurry were combined and incubated in a Thermomixer for 2 h at 30°C and 850 rpm to allow protein binding via Cys thiols.…”

“…We employed the data analysis pipeline described in detail in Berg et al (2019). The first step is the filtering of data features where all samples were outliers (Tukey's lower fence with k = 1.5).…”

“…We then stabilized the variance of datasets by performing a log 2 -tranformation of all abundance values. We then perform missing data imputation method using random sampling from a normal distribution with parameters robustly estimated from the entire dataset (mean value for the imputation is calculated per Cys for all nonmissing values in each condition, or using Tukey's lower fence with k = 2 if all values are missing in one condition; SD for the imputation is computed per condition by taking the median of SDs of each Cys with nonmissing values; Berg et al, 2019). We quantify relative changes in peptide abundance between conditions using the limma R package (Ritchie et al, 2015).…”

Proceedings of the 2018 MidSouth Computational Biology and Bioinformatics Society (MCBIOS) conference

“…Best Paper Award, MCBIOS 2018 : Phillip Berg et al, “Evaluation of Linear Models and Missing Value Imputation for the Analysis of Peptide-Centric Proteomics” [1].…”