Statistical Analysis of ATM-Dependent Signaling in Quantitative Mass Spectrometry Phosphoproteomics

Abstract

Ataxia-telangiectasia mutated (ATM) is a serine/threonine protein kinase, which when perturbed is associated with modified protein signaling that ultimately leads to a range of neurological and DNA repair defects. Recent advances in phospho-proteomics coupled with high-resolution mass-spectrometry provide new opportunities to dissect signaling pathways that ATM utilize under a number of conditions. This chapter begins by providing a brief overview of ATM function, its various regulatory roles and then leads into a workflow focused on the use of the statistical programming language R, together with code, for the identification of ATM-dependent substrates in the cytoplasm. This chapter cannot cover statistical properties in depth nor the range of possible methods in great detail, but instead aims to equip researchers with a set of tools to perform analysis between two conditions through examples with R functions.

The temporal profile of activity-dependent presynaptic phospho-signalling reveals long-lasting patterns of poststimulus regulation

“…Only high-confidence phosphopeptides with a probability score ≥ 0.75 for phosphorylation site assignment detected in at least three of the six biological replicates were used for further analyses. Our statistical workflow (Fig 1A and Materials and methods) involved normalization, missing value imputation, and correction for nonbiological sources of variation, as described [31, 32]. Unsupervised principle component analysis (PCA) of all phosphopeptides separated the time points across the first principle component, indicating that phosphorylation of peptides was largely occurring in a time-dependent manner (Fig 1B).…”

“…Initial inspection indicated requirement for normalization and correction of nonbiological sources of variation. For downstream normalization and analysis, we followed the statistical preprocessing method we previously implemented for analysis of phosphoproteomics data [31], and more detail is provided in Waardenberg, 2017 [32]. Data were first rescaled using quantile normalization, assuming that data were missing at random [76], followed by missing value imputation using the k-nearest neighbour approach (k = 10) [77], and data were iteratively reweighted using the surrogate variable analysis [78].…”

TWIST1 and chromatin regulatory proteins interact to guide neural crest cell differentiation

“…Processing and analysis of raw peptide-spectrum match (PSM) values were performed in R following the published protocol (Waardenberg, 2017). Data were normalized by the sum of PSM for each sample (Figure 1-figure supplement 2B), based on the assumption that the same amount of starting materials was loaded onto the mass spectrometer for the test and control samples.…”

TWIST1 and chromatin regulatory proteins interact to guide neural crest cell differentiation

“…Processing and analysis of raw peptide-spectrum match (PSM) values were performed in R following the published protocol (Waardenberg, 2017). Data were normalized by the sum of PSM for each sample ( Figure S2B), based on the assumption that the same amount of starting materials was loaded onto the mass spectrometer for the test and control samples.…”