Accounting for the Multiple Natures of Missing Values in Label-Free Quantitative Proteomics Data Sets to Compare Imputation Strategies

The impact of ambient air pollution on the human blood metabolome

BACKGROUND

Biological perturbations caused by air pollution might be reflected in the compounds present in blood originating from air pollutants and endogenous metabolites influenced by air pollution (defined here as part of the blood metabolome). We aimed to assess the perturbation of the blood metabolome in response to short term exposure to air pollution.

METHODS

We exposed 31 healthy volunteers to ambient air pollution for 5h. We measured exposure to particulate matter, particle number concentrations, absorbance, elemental/organic carbon, trace metals, secondary inorganic components, endotoxin content, gaseous pollutants, and particulate matter oxidative potential. We collected blood from the participants 2h before and 2 and 18h after exposure. We employed untargeted metabolite profiling to monitor 3873 metabolic features in 493 blood samples from these volunteers. We assessed lung function using spirometry and six acute phase proteins in peripheral blood. We assessed the association of the metabolic features with the measured air pollutants and with health markers that we previously observed to be associated with air pollution in this study.

RESULTS

We observed 89 robust associations between air pollutants and metabolic features two hours after exposure and 118 robust associations 18h after exposure. Some of the metabolic features that were associated with air pollutants were also associated with acute health effects, especially changes in forced expiratory volume in 1s. We successfully identified tyrosine, guanosine, and hypoxanthine among the associated features. Bioinformatics approach Mummichog predicted enriched pathway activity in eight pathways, among which tyrosine metabolism.

CONCLUSIONS

This study demonstrates for the first time the application of untargeted metabolite profiling to assess the impact of air pollution on the blood metabolome.

Using hyperLOPIT to perform high-resolution mapping of the spatial proteome

The organization of eukaryotic cells into distinct subcompartments is vital for all functional processes, and aberrant protein localization is a hallmark of many diseases. Microscopy methods, although powerful, are usually low-throughput and dependent on the availability of fluorescent fusion proteins or highly specific and sensitive antibodies. One method that provides a global picture of the cell is localization of organelle proteins by isotope tagging (LOPIT), which combines biochemical cell fractionation using density gradient ultracentrifugation with multiplexed quantitative proteomics mass spectrometry, allowing simultaneous determination of the steady-state distribution of hundreds of proteins within organelles. Proteins are assigned to organelles based on the similarity of their gradient distribution to those of well-annotated organelle marker proteins. We have substantially re-developed our original LOPIT protocol (published by Nature Protocols in 2006) to enable the subcellular localization of thousands of proteins per experiment (hyperLOPIT), including spatial resolution at the suborganelle and large protein complex level. This Protocol Extension article integrates all elements of the hyperLOPIT pipeline, including an additional enrichment strategy for chromatin, extended multiplexing capacity of isobaric mass tags, state-of-the-art mass spectrometry methods and multivariate machine-learning approaches for analysis of spatial proteomics data. We have also created an open-source infrastructure to support analysis of quantitative mass-spectrometry-based spatial proteomics data (http://bioconductor.org/packages/pRoloc) and an accompanying interactive visualization framework (http://www. bioconductor.org/packages/pRolocGUI). The procedure we outline here is applicable to any cell culture system and requires ∼1 week to complete sample preparation steps, ∼2 d for mass spectrometry data acquisition and 1-2 d for data analysis and downstream informatics.

Experimental design and data-analysis in label-free quantitative LC/MS proteomics: A tutorial with MSqRob

Label-free shotgun proteomics is routinely used to assess proteomes. However, extracting relevant information from the massive amounts of generated data remains difficult. This tutorial provides a strong foundation on analysis of quantitative proteomics data. We provide key statistical concepts that help researchers to design proteomics experiments and we showcase how to analyze quantitative proteomics data using our recent free and open-source R package MSqRob, which was developed to implement the peptide-level robust ridge regression method for relative protein quantification described by Goeminne et al. MSqRob can handle virtually any experimental proteomics design and outputs proteins ordered by statistical significance. Moreover, its graphical user interface and interactive diagnostic plots provide easy inspection and also detection of anomalies in the data and flaws in the data analysis, allowing deeper assessment of the validity of results and a critical review of the experimental design. Our tutorial discusses interactive preprocessing, data analysis and visualization of label-free MS-based quantitative proteomics experiments with simple and more complex designs. We provide well-documented scripts to run analyses in bash mode on GitHub, enabling the integration of MSqRob in automated pipelines on cluster environments (https://github.com/statOmics/MSqRob).

SIGNIFICANCE

The concepts outlined in this tutorial aid in designing better experiments and analyzing the resulting data more appropriately. The two case studies using the MSqRob graphical user interface will contribute to a wider adaptation of advanced peptide-based models, resulting in higher quality data analysis workflows and more reproducible results in the proteomics community. We also provide well-documented scripts for experienced users that aim at automating MSqRob on cluster environments.

From protein-protein interactions to protein co-expression networks: a new perspective to evaluate large-scale proteomic data

The reductionist approach of dissecting biological systems into their constituents has been successful in the first stage of the molecular biology to elucidate the chemical basis of several biological processes. This knowledge helped biologists to understand the complexity of the biological systems evidencing that most biological functions do not arise from individual molecules; thus, realizing that the emergent properties of the biological systems cannot be explained or be predicted by investigating individual molecules without taking into consideration their relations. Thanks to the improvement of the current -omics technologies and the increasing understanding of the molecular relationships, even more studies are evaluating the biological systems through approaches based on graph theory. Genomic and proteomic data are often combined with protein-protein interaction (PPI) networks whose structure is routinely analyzed by algorithms and tools to characterize hubs/bottlenecks and topological, functional, and disease modules. On the other hand, co-expression networks represent a complementary procedure that give the opportunity to evaluate at system level including organisms that lack information on PPIs. Based on these premises, we introduce the reader to the PPI and to the co-expression networks, including aspects of reconstruction and analysis. In particular, the new idea to evaluate large-scale proteomic data by means of co-expression networks will be discussed presenting some examples of application. Their use to infer biological knowledge will be shown, and a special attention will be devoted to the topological and module analysis.

A Bioconductor workflow for processing and analysing spatial proteomics data

Spatial proteomics is the systematic study of protein sub-cellular localisation. In this workflow, we describe the analysis of a typical quantitative mass spectrometry-based spatial proteomics experiment using the MSnbase and pRoloc Bioconductor package suite. To walk the user through the computational pipeline, we use a recently published experiment predicting protein sub-cellular localisation in pluripotent embryonic mouse stem cells. We describe the software infrastructure at hand, importing and processing data, quality control, sub-cellular marker definition, visualisation and interactive exploration. We then demonstrate the application and interpretation of statistical learning methods, including novelty detection using semi-supervised learning, classification, clustering and transfer learning and conclude the pipeline with data export. The workflow is aimed at beginners who are familiar with proteomics in general and spatial proteomics in particular.

A Bioconductor workflow for processing and analysing spatial proteomics data

Spatial proteomics is the systematic study of protein sub-cellular localisation. In this workflow, we describe the analysis of a typical quantitative mass spectrometry-based spatial proteomics experiment using the MSnbase and pRoloc Bioconductor package suite. To walk the user through the computational pipeline, we use a recently published experiment predicting protein sub-cellular localisation in pluripotent embryonic mouse stem cells. We describe the software infrastructure at hand, importing and processing data, quality control, sub-cellular marker definition, visualisation and interactive exploration. We then demonstrate the application and interpretation of statistical learning methods, including novelty detection using semi-supervised learning, classification, clustering and transfer learning and conclude the pipeline with data export. The workflow is aimed at beginners who are familiar with proteomics in general and spatial proteomics in particular.

Feature selection in clinical proteomics: with great power comes great reproducibility

In clinical proteomics, reproducible feature selection is unattainable given the standard statistical hypothesis-testing framework. This leads to irreproducible signatures with no diagnostic power. Instability stems from high P-value variability (p_var), which is inevitable and insolvable. The impact of p_var can be reduced via power increment, for example increasing sample size and measurement accuracy. However, these are not realistic solutions in practice. Instead, workarounds using existing data such as signal boosting transformation techniques and network-based statistical testing is more practical. Furthermore, it is useful to consider other metrics alongside P-values including confidence intervals, effect sizes and cross-validation accuracies to make informed inferences.

WGCNA Application to Proteomic and Metabolomic Data Analysis

Progresses in mass spectrometric instrumentation and bioinformatics identification algorithms made over the past decades allow quantitative measurements of relative or absolute protein/metabolite amounts in cells in a high-throughput manner, which has significantly expedited the exploration into functions and dynamics of complex biological systems. However, interpretation of high-throughput data is often restricted by the limited availability of suitable computational methods and enough statistical power. While many computational methodologies have been developed in the past decades to address the issue, it becomes clear that network-focused rather than individual gene/protein-focused strategies would be more appropriate to obtain a complete picture of cellular responses. Recently, an R analytical package named as weighted gene coexpression network analysis (WGCNA) was developed and applied to high-throughput microarray or RNA-seq datasets since it provides a systems-level insights, high sensitivity to low abundance, or small fold changes genes without any information loss. The approach was also recently applied to proteomic and metabolomic data analysis. However, due to the fact that low coverage of the current proteomic and metabolomic analytical technologies, causing the format of datasets are often incomplete, the method needs to be modified so that it can be properly utilized for meaningful biologically interpretation. In this chapter, we provide a detailed introduction of the modified protocol and its tutorials for applying the WGCNA approach in analyzing proteomic and metabolomic datasets.

Diastolic Left Ventricular Function in Relation to Urinary and Serum Collagen Biomarkers in a General Population

Current knowledge on the pathogenesis of diastolic heart failure predominantly rests on case-control studies involving symptomatic patients with preserved ejection fraction and relying on invasive diagnostic procedures including endomyocardial biopsy. Our objective was to gain insight in serum and urinary biomarkers reflecting collagen turnover and associated with asymptomatic diastolic LV dysfunction. We randomly recruited 782 Flemish (51.3% women; 50.5 years). We assessed diastolic LV function from the early and late diastolic peak velocities of the transmitral blood flow and of the mitral annulus. By sequencing urinary peptides, we identified 70 urinary collagen fragments. In serum, we measured carboxyterminal propeptide of procollagen type 1 (PICP) as marker of collagen I synthesis and tissue inhibitor of matrix metalloproteinase type 1 (TIMP-1), an inhibitor of collagen-degrading enzymes. In multivariable-adjusted analyses with Bonferroni correction, we expressed effect sizes per 1-SD in urinary collagen I (uCI) or collagen III (uCIII) fragments. In relation to uCI fragments, e’ decreased by 0.183 cm/s (95% confidence interval, 0.017 to 0.350; p = 0.025), whereas E/e’ increased by 0.210 (0.067 to 0.353; p = 0.0012). E/e’ decreased with uCIII by 0.168 (0.021 to 0.316; p = 0.018). Based on age-specific echocardiographic criteria, 182 participants (23.3%) had subclinical diastolic LV dysfunction. Partial least squares discriminant analysis contrasting normal vs. diastolic LV dysfunction confirmed the aforementioned associations with the uCI and uCIII fragments. PICP and TIMP-1 increased in relation to uCI (p<0.0001), whereas these serum markers decreased with uCIII (p≤0.0006). Diastolic LV dysfunction was associated with higher levels of TIMP-1 (653 vs. 696 ng/mL; p = 0.013). In a general population, the non-invasively assessed diastolic LV function correlated inversely with uCI and serum markers of collagen I deposition, but positively with uCIII. These observations generalise previous studies in patients to randomly recruited people, in whom diastolic LV function ranged from normal to subclinical impairment, but did not encompass overt diastolic heart failure.

The Primary Effect on the Proteome of ARID1A-mutated Ovarian Clear Cell Carcinoma is Downregulation of the Mevalonate Pathway at the Post-transcriptional Level

Inactivating mutations in ARID1A, which encodes a subunit of the SWI/SNF chromatin-remodeling complex, are found in over half of ovarian clear cell carcinoma cases and more broadly across most types of cancers. To identify ARID1A-dependent changes in intracellular signaling pathways, we performed proteome analyses of isogenic ovarian clear cell carcinoma cell lines with or without ARID1A expression. Knockout of ARID1A in an ovarian clear cell carcinoma cell line with wild-type ARID1A, OVCA429, primarily resulted in downregulation of the mevalonate pathway, an important metabolic pathway involved in isoprenoid synthesis, cholesterol synthesis, and other downstream pathways. In a complementary experiment, expression of wild-type ARID1A in an ovarian clear cell carcinoma cell line containing mutated ARID1A, OVISE, affected the mevalonate pathway in a reciprocal manner. A striking aspect of these analyses was that, although only 5% of the detected proteome showed significant abundance changes, most proteins in the mevalonate pathway were coordinately affected by ARID1A status. There were generally corresponding changes when comparing the proteomics data to our previously published microarray data for ectopic expression of ARID1A in the OVISE cell line. However, ARID1A-dependent changes were not detected for genes within the mevalonate pathway. This discrepancy suggests that the mevalonate pathway is not regulated directly by ARID1A-mediated transcription and may be regulated post-transcriptionally. We conclude that ARID1A status indirectly influences the mevalonate pathway and probably influences other processes including glycogen metabolism and 14-3-3-mediated signaling. Further, our findings demonstrate that changes in mRNA levels are sometimes poor indicators of signaling pathways affected by gene manipulations in cancer cells.