Jemma X. Wu

Rice crops are often subject to multiple abiotic stresses simultaneously in both natural and cultivated environments, resulting in yield reductions beyond those expected from single stress. We report physiological changes after a 4 day exposure to combined drought, salt and extreme temperature treatments, following a 2 day salinity pre-treatment in two rice genotypes—Nipponbare (a paddy rice) and IAC1131 (an upland landrace). Stomata closed after two days of combined stresses, causing intercellular CO2 concentrations and assimilation rates to diminish rapidly. Abscisic acid (ABA) levels increased at least five-fold but did not differ significantly between the genotypes. Tandem Mass Tag isotopic labelling quantitative proteomics revealed 6215 reproducibly identified proteins in mature leaves across the two genotypes and three time points (0, 2 and 4 days of stress). Of these, 987 were differentially expressed due to stress (cf. control plants), including 41 proteins that changed significantly in abundance in all stressed plants. Heat shock proteins, late embryogenesis abundant proteins and photosynthesis-related proteins were consistently responsive to stress in both Nipponbare and IAC1131. Remarkably, even after 2 days of stress there were almost six times fewer proteins differentially expressed in IAC1131 than Nipponbare. This contrast in the translational response to multiple stresses is consistent with the known tolerance of IAC1131 to dryland conditions.

In this protocol we describe our workflow for analyzing complex, multi-condition quantitative proteomic experiments, with the aim to extract biological insights. The tool we use is an R package, PloGO2, contributed to Bioconductor, which we can optionally precede by running correlation network analysis with WGCNA. We describe the data required and the steps we take, including detailed code examples and outputs explanation. The package was designed to generate gene ontology or pathway summaries for many data subsets at the same time, visualize protein abundance summaries for each biological category examined, help determine enriched protein subsets by comparing them all to a reference set, and suggest key highly correlated hub proteins, if the optional network analysis is employed. Key words Proteomics Pathway Gene ontology Functional enrichment analysis WGCNA Statistical R package

Credible detection and quantification of low abundance proteins from human blood plasma is a major challenge in precision medicine biomarker discovery when using mass spectrometry (MS). In this proof-of-concept study, we employed a mixture of selected recombinant proteins in DDA libraries to subsequently identify (not quantify) cancer-associated low abundance plasma proteins using SWATH/DIA. The exemplar DDA recombinant protein spectral library (rPSL) was derived from tryptic digestion of 36 recombinant human proteins that had been previously implicated as possible cancer biomarkers from both our own and other studies. The rPSL was then used to identify proteins from nondepleted colorectal cancer (CRC) EDTA plasmas by SWATH-MS. Most (32/36) of the proteins used in the rPSL were reliably identified from CRC plasma samples, including 8 proteins (i.e., BTC, CXCL10, IL1B, IL6, ITGB6, TGFα, TNF, TP53) not previously detected using high-stringency protein inference MS according to PeptideAtlas. The rPSL SWATH-MS protocol was compared to DDA-MS using MARS-depleted and postdigestion peptide fractionated plasmas (here referred to as a human plasma DDA library). Of the 32 proteins identified using rPSL SWATH, only 12 could be identified using DDA-MS. The 20 additional proteins exclusively identified using the rPSL SWATH approach were almost exclusively lower abundance (i.e., <10 ng/mL) proteins. To mitigate justified FDR concerns, and to replicate a more typical library creation approach, the DDA rPSL library was merged with a human plasma DDA library and SWATH identification repeated using such a merged library. The majority (33/36) of the low abundance plasma proteins added from the rPSL were still able to be identified using such a merged library when high-stringency HPP Guidelines v3.0 protein inference criteria were applied to our data set. The MS data set has been deposited to ProteomeXchange Consortium via the PRIDE partner repository (PXD022361).

We describe a useful workflow for characterizing proteomics experiments incorporating many conditions and abundance data, incorporating the popular approach weighted gene correlation network analysis (WGCNA) and functional annotation with the PloGO2 R package, the latter of which we have extended and made available to Bioconductor. The approach can use quantitative data from labelled or label-free experiments, and was developed to handle multiple files stemming from data partition or multiple pairwise comparisons. The WGCNA approach can similarly produce a potentially large number of clusters of interest, which can also be characterised functionally using PloGO2. Enrichment analysis will identify clusters or subsets of proteins of interest, and the WGCNA network topology scores will produce a ranking of proteins within these clusters or subsets. This can naturally lead to prioritized proteins to be considered for further analysis or as candidates of interest for validation in the context of complex experiments. We demonstrate the use of the package on two published datasets using 2 different biological systems (plant and human plasma) and proteomics platforms (SWATH and TMT); an analysis of the effect of drought on rice over time generated using TMT, and a paediatric plasma sample dataset generated using SWATH. In both, the automated workflow recapitulates key insights or observations of the published papers, and provides additional suggestions for further investigation. These findings indicate that dataset analysis using WGCNA combined with the updated PloGO2 package is a powerful method to gain biological insights from complex multi-faceted proteomics experiments.

Proteomics and genomics discovery experiments generate increasingly large result tables, necessitating more researcher time to convert the biological data into new knowledge. Literature review is an important step in this process and can be tedious for large scale experiments. An informed and strategic decision about which biomolecule targets should be pursued for follow-up experiments thus remains a considerable challenge. To streamline and formalise this process of literature retrieval and analysis of discovery based ‘omics data and as a decision-facilitating support tool for follow-up experiments we present OmixLitMiner, a package written in the computational language R. The tool automates the retrieval of literature from PubMed based on UniProt protein identifiers, gene names and their synonyms, combined with user defined contextual keyword search (i.e., gene ontology based). The search strategy is programmed to allow either strict or more lenient literature retrieval and the outputs are assigned to three categories describing how well characterized a regulated gene or protein is. The category helps to meet a decision, regarding which gene/protein follow-up experiments may be performed for gaining new knowledge and to exclude following already known biomarkers. We demonstrate the tool’s usefulness in this retrospective study assessing three cancer proteomics and one cancer genomics publication. Using the tool, we were able to corroborate most of the decisions in these papers as well as detect additional biomolecule leads that may be valuable for future research.

Post-translational modifications (PTMs) can occur soon after translation or at any stage in the lifecycle of a given protein, and they may help regulate protein folding, stability, cellular localisation, activity, or the interactions proteins have with other proteins or biomolecular species. PTMs are crucial to our functional understanding of biology, and new quantitative mass spectrometry (MS) and bioinformatics workflows are maturing both in labelled multiplexed and label-free techniques, offering increasing coverage and new opportunities to study human health and disease. Techniques such as Data Independent Acquisition (DIA) are emerging as promising approaches due to their re-mining capability. Many bioinformatics tools have been developed to support the analysis of PTMs by mass spectrometry, from prediction and identifying PTM site assignment, open searches enabling better mining of unassigned mass spectra—many of which likely harbour PTMs—through to understanding PTM associations and interactions. The remaining challenge lies in extracting functional information from clinically relevant PTM studies. This review focuses on canvassing the options and progress of PTM analysis for large quantitative studies, from choosing the platform, through to data analysis, with an emphasis on clinically relevant samples such as plasma and other body fluids, and well-established tools and options for data interpretation.