Pitfalls in phosphoproteomic data analysis

Phosphorylation of proteins plays a major role in signal transduction and provides insights into the activation status of pathways in response to stimuli. Investigation of global changes in post-translational modifications (PTMs) is best achieved using mass spectrometry (MS)-based proteomics. Various techniques for enriching phosphorylated peptides have advanced "phospho"-proteomics in recent years. Similar to global proteomics, the phospho-proteomics field is transitioning from data-dependent acquisition (DDA) combined with the MaxQuant/Perseus (MPI, Martinsried) analysis pipeline to data-independent acquisition (DIA) methods. Initially, DIA spectra were commonly analyzed using Spectronaut software (Biognosys), with precursor lists exported and converted into a Perseus-readable format for further data analysis. However, recent versions of Spectronaut include a generic phospho data analysis capability.

In this study, we compare data analysis in Perseus with Spectronaut using a real-life cell culture experiment. Phosphopeptides were enriched using Zr-IMAC beads (ReSyn) and measured in DIA (QE-HFX). Spectra were searched in directDIA mode in Spectronaut (version 17). Collapsed peptides were first analyzed in Perseus (version 2.0.3.1). Antibody-based validation of statistically significant differential peptides was performed using Western blots (WBs) for seven candidates. Subsequently, data analysis, including statistical evaluation, were tested in Spectronaut.

These datasets were systematically compared, focusing on localization probability cut-offs, imputation strategies, and different normalization options. Given that neither the data analysis parameters themselves, nor the order in which steps are performed are independent of each other, we aim to highlight potential pitfalls in data analysis and propose a best practice approach, guided by our positive control candidates.