The PRIDE database resources in 2022: a hub for mass spectrometry-based proteomics evidences

INTRODUCTION

Data sharing in the public domain has become the standard for proteomics researchers. The growth in recent years has been very remarkable and as a result, the number of proteomics datasets deposited every year in open public repositories is now comparable to transcriptomics (1). Since 2004, the PRoteomics IDEntifications (PRIDE) database (https://www.ebi.ac.uk/pride/) at the European Bioinformatics Institute (EMBL-EBI, Hinxton, Cambridge, UK) has enabled public data deposition of mass spectrometry (MS)-based proteomics data, providing access to the experimental data described in scientific publications (2). Since then, and especially in recent years, PRIDE Archive (the archival component of PRIDE) has become the largest repository for proteomics data sharing worldwide (2,3).

PRIDE stores datasets coming from all proteomics experimental approaches, with a focus on discovery-driven techniques such data dependent acquisition (DDA) and data independent acquisition (DIA) bottom-up proteomics, but also top-down proteomics and MS imaging, among others. For each dataset submitted to PRIDE Archive (the archival component of PRIDE), the MS raw files (output files from the mass spectrometers) and the processed results (at least peptide/protein identification results, quantification information is optional) must be provided. In addition, each dataset in PRIDE Archive can contain peptide/protein quantitation result files, the mass spectra as peak list files, the searched protein sequence databases or spectral libraries, programming scripts, and any other technical and/or biological metadata provided by the data submitters (4). The PRIDE team has led within the Proteomics Standards Initiative (PSI) organization, the creation and implementation of multiple standard open file formats such as mzTab (5), mzIdentML (6) and mzML (7) to store, process and visualize the proteomics data deposited.

The stand-alone ProteomeXchange (PX) Submission tool (8) allows the researchers to perform the data submissions to PRIDE Archive, while PRIDE Inspector (9) enables users to review the dataset before, during, and after has been deposited in the resource. After the submission is completed, different pipelines perform the validation and quality assessment of the reported results and store the data into multiple databases for enabling data access and visualization in the PRIDE Archive web interface (https://www.ebi.ac.uk/pride/archive) and also programmatically via the PRIDE Application Programming Interface (API, https://www.ebi.ac.uk/pride/ws/archive/v2/). In recent years, PRIDE Archive has been moving its visualization components from desktop-based applications (e.g., PRIDE Inspector) to Restful APIs and web-based interfaces. All submitted files are available to download via FTP or the Aspera file transfer protocol.

PRIDE resources have two main missions for the proteomics community: (i) support data deposition and quality assessment of submitted proteomics experiments, to help reproducible research; and (ii) promote and facilitate the reuse of public proteomics data, and disseminate high-quality proteomics evidences into added-value resources, including Ensembl (10), UniProt (11) and Expression Atlas (12).

The PRIDE database was one of the founders of the PX consortium in 2011 (3,8). PX defines the guidelines for data submission and dissemination of public proteomics data worldwide. As of 2021, the resources PeptideAtlas (13), including its related resource PASSEL (PeptideAtlas SRM Experiment Library) (14), MassIVE (15), jPOST (16), iProX (17) and Panorama Public (18) are the active members of the consortium. PX coordinates the release of accession numbers for every submitted dataset and a set of services for providing unified access to publicly available datasets (http://proteomecentral.proteomexchange.org/cgi/GetDataset), including specific data types such as mass spectra, using Universal Spectrum Identifiers (19) (http://proteomecentral.proteomexchange.org/usi/). Additionally, in 2017, PRIDE became an ELIXIR (http://www.elixir-europe.org) core data resource (20) and ELIXIR deposition database, recognizing its key role in the life sciences.

In this manuscript, we will summarize the main PRIDE-related developments in the last three years, since the previous Nucleic Acids Research (NAR) database update manuscript was published (2). We will discuss PRIDE Archive first but will also provide updated information about the PRIDE-related tools and other ongoing activities including the updates in the PRIDE Spectra Archive and PRIDE Peptidome. Additionally, we will also report about the work performed to disseminate and integrate proteomics data in other EMBL-EBI resources.

CURRENT STATUS OF THE PRIDE ECOSYSTEM: RESOURCES AND TOOLS

The PRIDE database ecosystem (https://www.ebi.ac.uk/pride/) is composed of a comprehensive set of libraries, desktop tools, databases, large-scale pipelines, Restful APIs and web applications (Figure 1). A set of open-source Java libraries including jmzTab (21), jmzIdentML (22), ms-data-core-api (23) and the protein inference algorithms toolkit (PIA) (24,25) supported and maintained by the PRIDE team, allows to read, validate, process, and store proteomics data encoded in PSI open file formats. PRIDE Archive pipelines (2) perform a set of validation and quality checks to make sure the deposited files are semantically valid, and that the metadata provided during the submission is correct, in addition to moving the submitted datasets into the EMBL-EBI production filesystem.

When a given dataset is made public, a group of post-submission pipelines parses the peptides and proteins identified in the dataset—if the dataset is a ‘complete’ submission (4)—and index them into Apache Solr and MongoDB-based infrastructure enabling to search datasets by the identified peptides and proteins. The PRIDE Spectra Archive and PRIDE Peptidome provide access to the mass spectra identified in the PRIDE Archive and to a condensed view of high-quality identified peptides across PRIDE Archive datasets, respectively. All data from PRIDE Archive and related resources are served through the PRIDE Restful API and the web application.

PRIDE ARCHIVE SUBMISSION STATISTICS

As of 1 August 2021, PRIDE Archive stored 23 168 datasets—compared to the 10 100 datasets available on August 2018 (2)—, which means that 56.4% of the data in PRIDE Archive has been submitted in the last 3 years. Figure 4 shows the distribution of submissions by month, species, and disease in PRIDE Archive since 2012, and the cumulative size of PRIDE Archive data in terabytes.

In 2019, PRIDE Archive received 314 datasets per month on average, 436 during 2020, and so far in 2021, this number has grown to 499 datasets on average (Figure 4A), which affirms the increasing huge demand and growing prominence of PRIDE. At the time of writing, PRIDE hosts ∼83% of all PX datasets, coming from >8 000 research groups, from 66 countries. The number of submitted datasets that are now publicly available is currently 64%, reflecting an improvement of around 8% when compared with 2019. With this aim in mind, the team has developed multiple mechanisms to detect datasets already published that have not been reported to PRIDE by the original submitters. As a concrete example, submitters can report via the PRIDE web interface datasets that have already a corresponding manuscript published, if the dataset is still private. The size of PRIDE Archive data has doubled from 2019 to 2021 (Figure 4B). As a result, PRIDE Archive is the third-largest omics Archive at EMBL-EBI only exceeded by the genomics resources ENA (European Nucleotide Archive) and EGA (European Genome-phenome Archive) (41).

As of September 2021, the majority of data in PRIDE Archive (including both public and private datasets) are human datasets (including cell lines) (39.1%), followed by mouse (13.7%), Saccharomyces cerevisiae (2.8%), Arabidopsis thaliana (2.7%), Rattus norvegicus (2.5%) and Escherichia coli (2.3%). Whereas most of the datasets come from model organisms, overall, datasets coming from >3 224 different taxonomy identifiers are stored in PRIDE Archive (Supplementary File S1).

The number of submitted datasets split by tissues and diseases are more heterogeneous (Figure 4C and D), being ‘cell-culture (non-specific tissue)’, and ‘disease-free (healthy/normal samples)’ the most predominant annotations. Altogether, cancer is the most studied disease followed by Alzheimer's and Parkinson's disease. Importantly, as of September 2021, more than 180 COVID-19 related datasets have been submitted to PRIDE Archive. These datasets, once they become publicly available, are integrated into the EMBL-EBI resource COVID-19 Data Portal (https://www.covid19dataportal.org/), enabling researchers to access all public data at EMBL-EBI resources in a unified interface (42).

PRIDE ARCHIVE AS A HUB OF MS EVIDENCES

Proteomics researchers are increasingly reusing public data from PRIDE (and other PX resources) for a broad range of purposes. For instance, recent resources that have been started by reusing mostly PRIDE public datasets include OpenProt (43), MatrisomeDB (44), Scop3P (45) and ProteomeHD (46). Additionally, as just one among many examples of high-profile data reuse, PRIDE datasets are routinely reanalyzed in the context of the Human Proteome Project (47). Figure 5A shows the increase in volumes of data downloaded from PRIDE Archive since 2013. Recently, PRIDE has started to track the reuse of public PRIDE datasets in publications. This information (if applicable) is available in the dataset web page when clicking on the term ‘Dataset reuses’. Figure 5B shows the increase in manuscripts (including pre-prints) published per year, where PRIDE datasets are reused.

Rather than in the creation of new resources, for sustainability reasons, our focus in-house has been put in disseminating and integrating PRIDE proteomics data into added-value EMBL-EBI resources such as UniProt (11), Ensembl (10), and Expression Atlas (12). Additionally, we have just started in the first steps of the work required to disseminate and integrate metaproteomics data into MGnify (48), an EMBL-EBI resource for the analysis, archiving, and browsing of metagenomic and metatranscriptomic data. The dissemination of public proteomics data into different resources has different goals depending on each specific resource but can be grouped in three main categories: (i) provide aggregated peptide/protein evidences as originally submitted to PRIDE Archive, in the case of UniProt and Ensembl; (ii) provide peptide/protein evidences, variant sequences and PTM information from reanalyzed datasets to UniProt, Ensembl and in the near future, to MGNify. In this case, an open analysis pipeline is used, including well-defined quality control metrics; and (iii) provide quantitative protein expression information into Expression Atlas, using data coming from reanalyzed datasets.

DISCUSSION AND FUTURE PLANS

Data deposition and dissemination have changed the proteomics community since the creation of PX almost 10 years ago. Most of the proteomics journals require nowadays the authors to deposit their data in a PX resource, which has enabled a better reproducibility and traceability of the claims reported in a given manuscript. The proteomics community is now widely embracing open data policies, an opposite scenario to the situation just a few years ago. At the same time, public proteomics data are being increasingly reused with multiple applications (1). We next outline some of the main working areas for PRIDE in the near future.

First of all, PRIDE is raising the bar of metadata annotation for all submitted datasets. MAGE-TAB for proteomics has been created with the aim that every submitted dataset provides information about the sample and the experimental design. The improvement in the annotation is also required to facilitate further data reuse for third parties. We expect that, gradually, the SDRF-Proteomics component will be made required for every dataset submission, after the community understands and get a full idea of the file format and of the mandatory information that needs to be provided. Multiple materials (https://github.com/bigbio/proteomics-metadata-standard/wiki), including examples and video tutorials, have been made available to better understand the file format and how it can be submitted to PRIDE Archive.

With the growing importance of clinical proteomics, i.e. in the context of multi-omics studies, another important area is the management of clinical sensitive human proteomics data. Ethical issues in proteomics are starting to be discussed and becoming increasingly relevant. A community-driven white paper on the topic has been recently published describing the current state-of-the-art (54). Addressing ethical issues for genomics and transcriptomics data led to processes to control who may access the data, so-called ‘controlled access’. Resources supporting the storage and dissemination of controlled access DNA/RNA sequencing datasets include the EGA and others internationally such as dbGAP (USA) and the Japanese Genotype-phenotype Archive. At present, all data in PRIDE (and in all PX resources) is fully open. Therefore, there is an increasing number of clinical sensitive human datasets that cannot be made available via PRIDE due to ethical-related issues (55). To address this problem, we will be working in developing a tailored infrastructure for sensitive human proteomics data, and in all the related policies. Additionally, in the context of data archiving activities, we plan to improve the support for cross-linking data - as outlined here (56) - and to provide better data integration for structural proteomics datasets between PRIDE Archive and the Protein Data Bank (PDB).

As shown above, we are already working on developing open and reproducible data analysis pipelines for different flavours of proteomics workflows (e.g., DDA, DIA, proteogenomics) (49,53,57). The main rationale is to make possible the use of that software in cloud infrastructures so that in the future the pipelines can be used by the community in the cloud using software container technologies (58). In addition, we aim to increasingly perform in-house data reuse (including data re-analysis) and disseminate high-quality proteomics data from PRIDE into the already mentioned added-value resources (Ensembl, UniProt, Expression Atlas, and MGnify in the near future). In this context, we will also work in improving the PRIDE Archive infrastructure to store dataset reanalyses appropriately, linking them to the relevant resources. One aim is to further develop data dissemination and integration practices also involving resources outside of EMBL-EBI.

To finalize, we invite interested parties in PRIDE-related developments to follow the PRIDE Twitter account (@pride_ebi). For regular announcements of all the new publicly available datasets, users can follow the PX Twitter account (@proteomexchange).

Supplementary Material

SUPPLEMENTARY DATA

Supplementary Data are available at NAR Online.

FUNDING

Wellcome [208391/Z/17/Z]; BBSRC grants ‘Proteomics DIA’ [BB/P024599/1], ‘PTMeXchange’ [BB/S01781X/1], ‘GRAPPA’ [BB/T019670/1]; UK-Japan Partnership award [BB/N022440/1]; NIH ‘Proteomics Standards’ grant [R24 GM127667-01]; EU H2020 project EPIC-XS [823839]; Open Targets [OTAR-043]; Luxembourg National Research Fund [C19/BM/13684739]; several ELIXIR Implementation Studies and EMBL core funding; M.E. and A.F.-Z. would like to acknowledge funding from de.NBI, a project of the German Federal Ministry of Education and Research (BMBF) [FKZ 031 A 534A]; Center for Protein Diagnostics (PPRODI), a grant of the Ministry of Innovation, Science and Research of North-Rhine Westphalia, Germany. Funding for open access charge: Wellcome.

Conflict of interest statement. None declared.