UniProt: the universal protein knowledgebase in 2021 — TEI-tagged view

Alex BatemanpersName, María MartinpersName, Sandra OrchardpersName, Michele MagranepersName, Rahat AgivetovapersName, Shadab AhmadpersName, Emanuele AlpipersName, Emily Bowler-BarnettpersName, Ramona BrittopersName, Borisas BursteinaspersName, Hema Bye‐A‐JeepersName, Ray CoetzeepersName, Austra CukurapersName, Alan Da SilvapersName, Paul DennypersName, Tunca DoğanpersName, ThankGod E. EbenezerpersName, Jun FanpersName, Leyla Jael CastropersName, Penelope GarmiripersName, George P. GeorghioupersName, Leonardo Jose da Costa GonzalespersName, Emma Hatton-EllispersName, Abdulrahman HusseinpersName, Alexandr IgnatchenkopersName, Giuseppe InsanapersName, Rizwan IshtiaqpersName, Petteri JokinenpersName, Vishal JoshipersName, Dushyanth JyothipersName, Antonia LockpersName, Rodrigo LópezpersName, Aurélien LucianipersName, Jie LuopersName, Yvonne LussipersName, Alistair MacDougallpersName, Fábio MadeirapersName, Mahdi MahmoudypersName, M. MenchipersName, Alok MishrapersName, Katie MoulangpersName, Andrew NightingalepersName, Carla Susana OliveirapersName, Sangya PundirpersName, Guoying QipersName, Shriya RajpersName, Daniel L RicepersName, M. Rodríguez-LópezpersName, Rabie SaidipersName, J. H. SampsonpersName, Tony SawfordpersName, Elena SperettapersName, E. B. TurnerpersName, Nidhi TyagipersName, Preethi VasudevpersName, Vladimir VolynkinpersName, Kate WarnerpersName, Xavier WatkinspersName, Rossana ZarupersName, Hermann ZellnerpersName, Alan BridgepersName, Sylvain PouxpersName, Nicole RedaschipersName, Lucila AimopersName, Ghislaine Argoud‐PuypersName, Andrea H AuchinclosspersName, Kristian B. AxelsenpersName, Parit BansalpersName, Delphine BaratinpersName, Marie-Claude BlatterpersName, Jerven BollemanpersName, Emmanuel BoutetpersName, Lionel BreuzapersName, Cristina Casals‐CasaspersName, Edouard de CastropersName, Kamal Chikh EchioukhpersName, Elisabeth CoudertpersName, Beatrice CuchepersName, Mikael DochepersName, Dolnide DornevilpersName, Anne EstreicherpersName, Maria Livia FamigliettipersName, Marc FeuermannpersName, Elisabeth GasteigerpersName, Sébastien GéhantpersName, Vivienne Baillie GerritsenpersName, Arnaud GospersName, Nadine Gruaz-GumowskipersName, Ursula HinzpersName, Chantal HulopersName, Nevila Hyka‐NouspikelpersName, Florence JungopersName, G. KellerpersName, Arnaud KerhornoupersName, V. LarapersName, Philippe Le MercierpersName, Damien LieberherrpersName, Thierry LombardotpersName, Xavier MartínpersName, Patrick MassonpersName, Anne MorgatpersName, Teresa Batista NetopersName, Salvo PaesanopersName, Ivo PedruzzipersName, Sandrine PilboutpersName, Lucille PourcelpersName, Monica PozzatopersName, Manuela PruesspersName, Catherine RivoirepersName, Christian J A SigristpersName, Karin SonessonpersName, André StutzpersName, Shyamala SundarampersName, Michael TognollipersName, Laure VerbreguepersName, Cathy WupersName, Cecilia N. ArighipersName, Leslie ArminskipersName, Chuming ChenpersName, Yongxing ChenpersName, John S. GaravellipersName, Hongzhan HuangpersName, Kati LaihopersName, Peter B. McGarveypersName, Darren A. NatalepersName, Karen RosspersName, C R VinayakapersName, Qinghua WangpersName, Yuqi WangpersName, Lai-Su YehpersName, Jian ZhangpersName, Patrick RuchpersName, Douglas TeodoropersName (2020date). “UniProt: the universal protein knowledgebase in 2021.”

Object of study — A KNOWLEDGEBASE THAT DELETES FROM ITSELF AND KEEPS WHAT IT DELETED - 190 million protein sequences in two provenance tiers, with a separate archive holding everything the knowledgebase has dropped

Research question — what UniProt now contains, how completeness and quality are assessed, and how annotations are evidenced

Contribution — a two-year status report: proteometerm completeness metrics, automated annotation, a credit-based community submissionterm route, and a COVID-19 curation portal

Full text — tagged

The reference list is not part of the text we hold for this paper. Our copy was extracted from the publisher's page and the bibliography did not survive that extraction, so the article below ends before its references. Nothing has been removed by us.

Abstract. Abstract The aim of the UniProt Knowledgebaseterm is to provide users with a comprehensive, high-quality and freely accessible set of protein sequencesterm annotated with functional informationterm. In this article, we describe significant updates that we have made over the last two years to the resource. The number of sequences in UniProtKB has risen to approximately 190 million, despite continued work to reduce sequence redundancyterm at the proteometerm level. We have adopted new methods of assessing proteome completeness and qualityterm. We continue to extract detailed annotations from the literature to add to reviewed entriesterm and supplement these in unreviewed entriesterm with annotations provided by automated systemsterm such as the newly implemented Association-Rule-Based Annotator (ARBA)term. We have developed a credit-based publication submission interfaceterm to allow the community to contribute publications and annotations to UniProt entries. We describe how UniProtKB responded to the COVID-19 pandemic through expert curationterm of relevant entries that were rapidly made available to the research communityterm through a dedicated portalterm. UniProt resources are available under a CC-BY (4.0) license via the web at https://www.uniprot.org/ . INTRODUCTION The UniProt databases exist to support biological and biomedical research by providing a complete compendium of all known protein sequence dataterm linked to a summary of the experimentally verifiedterm, or computationally predictedterm, functional informationterm about that protein. The UniProt Knowledgebaseterm (UniProtKB) combines reviewed UniProtKB/Swiss-Prot entriesterm, to which data have been added by our expert biocuration teamterm, with the unreviewed UniProtKB/TrEMBL entriesterm that are annotated by automated systemsterm. The UniRef databasesterm cluster sequence sets at various levels of sequence identityterm and the UniProt Archiveterm (UniParcterm) delivers a complete set of known sequences, including historical obsolete sequencesterm. UniProt additionally integrates, interprets, and standardizes data from multiple selected resources to add biological knowledge and associated metadata to protein records and acts as a central hubterm from which users can link out to 180 other resourcesterm. In recognition of the quality of our data, and the service we provide, UniProt was recognised as an ELIXIR Core Data Resource in 2017 ( 1 ) and received the CoreTrustSeal certification in 2020. The data resource fully supports the Findable, Accessible, Interoperable and Reusableterm (FAIRterm) data principles ( 2 ), for example by making data available in a number of community recognised formats, such as text, XML and RDF and via Application Programming Interfacesterm (API)s and File Transfer Protocol (FTP) downloadsterm, providing stable and traceable identifiersterm for protein sequence and protein sequence features and by fully evidencing our data sources throughout. We have also reviewed and updated our data licencing policies. UniProt is continually evolving to meet new challenges while still working to capture all available protein sequence data and to curate the ever-increasing amount of functional data described in the scientific literature. In our last update published in this journal in 2019 ( 3 ), we described how we are responding to the growth in microbial protein sequence records, largely derived from high-quality metagenomic assembled genomes. These will increasingly be added to by large-scale eukaryotic sequencing programs, such as the Darwin Tree of Life ( www.darwintreeoflife.org ) and Earth Biogenome ( www.earthbiogenome.org ) projects. Collectively, these have already resulted in the number of entries contained in UniProtKB growing by >65 million records, an increase of >50% in just 2 years. As the volume of sequence data continues to grow, we will continue to explore different ways to ensure database sustainability and scalability whilst still providing the best possible service to our user community. PROGRESS AND NEW DEVELOPMENTS Growth of sequence records in UniProt UniProt release 2020_04 contains over 189 million sequence records (Figure 1 ), with >292 000 proteomesterm, the complete set of proteins believed to be expressed by an organismterm, originating from completely sequenced viral, bacterial, archaeal and eukaryotic genomes available through the UniProtKB Proteomesterm portal ( https://www.uniprot.org/proteomesterm/ ). The majority of these proteomesterm continue to be based on the translation of genome sequence submissions to the INSDC source databasesterm—ENA, GenBank and the DDBJ ( 4 )—supplemented by genomes sequenced and/or annotated by groups such as Ensembl ( 5 ), NCBI RefSeq ( 6 ), Vectorbase ( 7 ) and WormBase ParaSite ( 8 ). Viral proteomesterm are manually checked and verified and periodically added to the database. Figure 1. Growth in the number of entries in the UniProt databases over the last decade. UniProt proteomesterm provide the set of proteins currently believed to be expressed by an organism. Previously, this dataset only consisted of complete proteomesterm derived from fully sequenced genomes. This has now been updated to enable the user to access a wider set of proteomesterm with different quality and completeness metricsterm. To enable researchers to evaluate proteometerm completeness and expected gene content, we have adopted the BUSCOterm (Benchmarking Universal Single-Copy Orthologsterm) scoring method for vertebrate, arthropod, fungal, and prokaryotic organisms on the Proteomesterm portal, in addition to providing details of species and the protein count. BUSCOterm v3 ( 9 ) identifies complete, duplicated, fragmented, and potentially missing genesterm by comparison to a defined set of near-universal single copy orthologsterm. As a result of our adoption of the BUSCOterm algorithm, a number of incomplete and poor-quality proteomesterm were identified and removed from UniProtKB. The Proteomesterm webpage has been redesigned to enable users to view full details of their proteometerm(s) of interest in a single table view (Figure 2 ). We also display the results of the ‘Complete Proteome Detectorterm’ (CPD), an in-house algorithm, which statistically evaluates the completeness and quality of each proteometerm by directly comparing it to those of a group of at least three closely taxonomically related species. The CPD classifies each proteometerm as either ‘standardterm’, ‘close to standardterm’ or an ‘outlierterm’, according to protein count vs. the standardterm distribution of protein count expected for completeness in comparison to a cluster of closely related organisms ( https://www.uniprot.org/help/assessing_proteomesterm ). We additionally provide the assessment of the genome assembly status imported from the source of the genome assembly and annotation (e.g. Ensembl or RefSeq). Users may sort by BUSCOterm or CPD score and/or filter by membership of the Reference Proteome setterm (∼7% of total proteomesterm) A Reference Proteometerm is one that has been selected either by the research communityterm or by computational clusteringterm ( 10 ) as providing the best annotated proteometerm in their cluster and is stably maintained as the chosen representative unless a higher quality proteometerm is identified in that cluster. UniProt Proteometerm pages now also provide a link to download a one-to-one protein set for the corresponding number of unique genes found in the genome. For each gene, a single protein sequence is algorithmically chosen from the proteometerm. This allows users to get a gene-centric subset of representative proteinsterm for a given genome, as opposed to the full proteometerm which includes all proteins (e.g. including isoformsterm) that map to the genome. Figure 2. Bacillus subtilis proteomesterm viewed on the Proteomesterm webpage with BUSCOterm and CPD scores. The left-hand panel suggests further option by which the user could filter the data, for example by only selecting reference proteomesterm. As previously described ( 3 ), we continue to remove almost identical proteomesterm of the same species from UniProtKB ( https://www.uniprot.org/help/proteometerm_redundancy ), leaving only a relevant representative proteometerm in that section of the database. The redundant proteome sequencesterm are available through UniParcterm to researchers and stable proteome identifiersterm (of the form UPXXXXXXXXX, where Xs are integers) are maintained for each redundant proteometerm to ensure findabilityterm. UniRef protein sequence clustersterm facilitate sequence similarity searchesterm, functional annotation, gene predictionterm, and genome and proteometerm comparisons. We have adopted the MMseqs2 algorithmterm to improve the speed of UniRef production ( 11 ), decreasing the time taken to perform UniRef50 clusteringterm from four weeks to 60 hours, and improved procedures to compute proteometerm clusters and identify representative proteomesterm, pan proteomesterm, and core and accessory proteomesterm. Expert curationterm The evaluation of experimental dataterm published in the scientific literature, and summarizing key points of biological relevance in the appropriate reviewed UniProtKB/Swiss-Prot record, is fundamental to the operation of the UniProt database. The functional informationterm extracted from the literature is added both in the form of human readable summariesterm and via structured vocabulariesterm, such as the Gene Ontology (GO)term ( 12 ). The curators also work to improve the computational accessibility of UniProt records, for example updating and replacing the existing textual descriptions of biochemical reactionsterm in UniProtKB using the Rhea knowledgebaseterm of biochemical reactionsterm ( 13 ). Rhea ( 14 ) uses the chemical ontology ChEBIterm (Chemical Entities of Biological Interest) ( 15 ) to describe reaction participantsterm, their chemical structuresterm and chemical transformationsterm. The complete ChEBI ontology is indexed to support hierarchical searchesterm on the UniProt website so that a user searching on a top-level term such as ‘phosphatidylinositol’ (CHEBI:28874) will find reactions involving derivatives such as 1-phosphatidyl-1D-myo-inositol 3,4,5-trisphosphate (CHEBI:16618). Users can also search using Rhea identifiers as well as identifiers, names, synonyms and chemical structuresterm (encoded as InChIKeysterm) from ChEBI. A user-friendly visualization for chemical reactions has also been developed to enable Rhea reactions to be viewed on the website within the appropriate UniProt entry (Figure 3 ). UniProtKB/Swiss-Prot currently includes annotations for 8,058 unique Rhea reactions, which feature in 220 003 distinct UniProtKB/Swiss-Prot protein records (39.1% of all UniProtKB/Swiss-Prot records are annotated with Rhea) (release 2020_04 of 12 August 2020). The annotation of pseudoenzymesterm has also been reviewed and updated, in collaboration with experts in this field ( 16 ). Figure 3. Catalytic activity comment and Rhea reaction visualized by the reaction graphic for the histone-lysine N -methyltransferase EHMT2 (UniProtKB: Q96KQ7 ). To further ensure our data is both human-readable and also computationally-tractable and continues to adhere to the FAIRterm principles ( 2 ), we are working to standardize the representation of all existing UniProt data on the functional impact of human variationterm. UniProtKB/Swiss-Prot has curated over 81 000 variants in 13,000 human protein sequence records from peer-reviewed literature, annotated with information on the functional impact and clinical significanceterm, when known. Over 30 000 of these variants have been associated with Mendelian diseasesterm. Clinical significanceterm is evaluated using the guidelines of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology (ACMG-AMPterm) ( 17 ) and ClinGen tools such as the pathogenicity calculatorterm ( 18 ), with all clinical interpretations routinely submitted to ClinVarterm to promote reuse ( 19 ). In addition to the increased use of structured vocabulariesterm to enhance accessibility to UniProtKB records, we have also improved the presentation of the information within each entry. The vast majority of multi-exon genes undergo alternative splicingterm to produce a variety of splice isoform proteinsterm, which can potentially increase the functional diversity of proteinsterm. Currently at least 95% of human genes are believed to be alternatively spliced ( 20 , 21 ) resulting in an estimated 75 000 distinct protein coding sequences. Post-translational proteolytic cleavageterm, where proteins are cleaved to remove some additional amino acid(s) or portion of protein, creates yet more mature amino-acid chainsterm as a single polyproteinterm may generate multiple bioactive proteins or peptidesterm. We have adapted our data model to capture machine-readable functional annotationsterm for specific isoformsterm and polyprotein cleavage productsterm, and now provide such knowledge for >5000 protein sequence entries. This enables the users to mine the data to identify cases where alternative protein sequences generated from the same gene have different functions. UniProtKB provides stable identifiers to both isoformsterm and to post-translationally cleaved chains which are used by many of our collaborators to identify these specific protein sequences in their own resources. This, in turn, has allowed us to further improve our mapping of imported data, for example binary protein interactionsterm imported from the IMEx Consortiumterm of molecular interaction databases are now displayed at the specific isoformterm/post-processed chain level. The representation of isoform-specific annotations and sequence features has been enhanced in the website to facilitate the exploration of this information in UniProt (Figure 4 ). Figure 4. Information extracted from an entry describing Hepatitis C viral protein (UniProtKB: P27958 ) highlighting annotation added at the processed mature chain level, describing the p21 core protein. (PRO_0000037566). Responding to the COVID-19 pandemic The broad expert knowledge of our curation team enables us to quickly respond to areas of new and rapidly expanding biology. Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) was identified as the cause of the 2019–2020 COVID-19 viral outbreak and ensuing pandemic. Its genome sequence was first made publicly available by the International Nucleotide Sequence Database Collaboration (INSDC) on 10 January 2020 (accession number MN908947 ). UniProt curators specializing in viral proteomesterm rapidly annotated the proteins encoded by this viral genome, first by similarity to other closely related coronaviruses, and subsequently by updating relevant entries with experimental dataterm as soon as this was published. A pre-release datasetterm was made publicly available, first as text files on the UniProt FTP site, followed by the launch of a dedicated COVID-19 disease portalterm in March 2020 ( https://covid-19.uniprot.org ), providing the latest available pre-release UniProtKB data for the SARS-CoV-2 coronavirus and other viral and human entries relating to the COVID-19 outbreak. As the pandemic has progressed, these entries have been prioritized for further update to support global research efforts, with information relating to a more granular understanding of protein functionalityterm, 3D-structureterm, binary interactions, and complex formationterm being added when either published or deposited in collaborating resources. The COVID portalterm is currently being updated more frequently than the standard 8-weekly UniProt release cycleterm to ensure the research communityterm accesses these data in a timely manner. The COVID-19 disease portalterm is a prototype of our intention to provide disease-centric access pointsterm to the wealth of data contained in UniProtKB records. Another recent example of a disease-focused curation effort has been the update of records relating to Alzheimer's diseaseterm, including proteins containing a disease-related amino-acid variantterm, their interacting partners and model organism proteins important for our understanding of diseaseterm initiation and progression. This work has been supported by clinical researchers active in the field who contributed to a number of workshops held in both the USA and UK and have suggested key protein targets for focused curation ( 22 , 23 ) and provided valuable user input into how this data should be accessed. Automatic annotationterm The ever-increasing amount of genomic data arising from current sequencing projects means that the proportion of unreviewed records in UniProtKB/TrEMBL describing largely predicted proteins represents by far the largest, and most rapidly growing, section of UniProtKB. These unreviewed records are enriched with functional annotation by systems using the protein classification tool InterProterm ( 24 ), which classifies sequences at superfamily, family and subfamily levelsterm, and predicts the occurrence of functional domainsterm and important sitesterm. In UniProtKB/TrEMBL entries, domains predicted by the InterProterm member databases PROSITE, SMART or Pfam are used to automatically provide domain annotations. The semi-automated rule-based computational annotation UniRule systemterm ( 25 ) annotates experimentally uncharacterized proteinsterm based on similarity to known experimentally characterized proteinsterm, adding properties, such as protein name, functional annotation, catalytic activity, pathway, GO terms and subcellular location. We continue to increase the number of UniRules used for annotation and this set has now grown to 6768 (release 2020_04) rules in total. To complement the expert guided process of creating UniRules, we have recently (release 2020_04) introduced the Association-Rule-Based Annotator (ARBA)term, a multiclass, self-training annotation systemterm for automatic classification and annotation of UniProtKB proteins ( 3 ). This replaces the previous rule-based SAAS systemterm. ARBAterm is trained on UniProtKB/Swiss-Prot, then uses rule mining techniquesterm to generate concise annotation modelsterm with the highest representativeness and coverage based on the properties of InterProterm group membership and taxonomy. ARBAterm employs a data exclusion setterm that censors data not suitable for computational annotation (such as specific biophysical or chemical properties) and generates human-readable rulesterm for each release which are made available at https://www.uniprot.org/arbaterm/ . 22 894 ARBA rulesterm were used to annotate 87 325 890 proteins in release 2020_04, increasing the combined coverage of the rule-based annotation systemsterm from 35% to 49% in UniProtKB/TrEMBL. Sequence feature predictions are currently excluded from annotation by ARBAterm. Additionally, in release 2020_04, more than 15 million uncharacterized protein namesterm have been improved using InterProterm member database signatures, updating their name to ‘domain X containing proteinterm’ following the International Protein Nomenclature Guidelines ( https://www.uniprot.org/docs/International_Protein_Nomenclature_Guidelines.pdf ). For example, the uncharacterized Western lowland gorilla protein UniProtKB:G3RLC3 has now been renamed ‘SH2 domain-containing protein’ giving biological information to the user. This system includes adding names based on domains of unknown function (e.g. UniProtKB:A0A009EMH9 DUF4372 domain-containing protein) as, although not immediately informative, it enables protein grouping thus improving the chances of eventually assigning a function to that domain. All automatic annotationsterm are labelled with their evidence/source. As previously described ( 3 ), UniFIREterm is an open-source Java-based framework and tool developed to apply the UniProt annotation rules on given protein sequences and provided by UniProt to share our knowledge in computational annotation and our rule-based systems ( https://gitlab.ebi.ac.uk/uniprot-public/unifireterm ). This system is freely available for groups to use for in-house protein annotation projects ( 26 ) or to contribute their own rules in the URML (UniProt Rule Markup Languageterm) format which may be reused for the annotation of UniProtKB entries. The automatic annotationterm systems described above require the presence of an ordered region of protein that can be recognized as a domain or provide a signature of family membership which has been identified by an InterProterm member database. However, over 20% of unreviewed proteins in UniProt do not contain any InterProterm signature regions, and many InterProterm signatures are not associated with transferable annotation. The largest part of missing annotation seems to derive from intrinsically disordered (ID) protein regionsterm, therefore we have collaborated with the MobiDB-lite resource to provide a consensus-based prediction of long disorderterm ( 27 ). The method uses eight different predictorsterm to derive a consensus which is then filtered for spurious short predictions. 109 144 661 predictions of regions of disorder, plus those described as ‘Basic’, ‘Polar’, ‘Acidic’, ‘Polyampholyte’ and ‘Pro-’ or ’Cys-rich’ have been added to 37 286 893 unreviewed entriesterm and it is planned to also import these annotations into the appropriate UniProtKB/Swiss-Prot entries. Data Integration UniProtKB integrates large-scale datasetsterm, mapping these data onto the appropriate protein sequence records and displaying the mappings via the ProtVista visualisation toolterm ( 28 ) and downloadable via FTP and APIs ( 29 ). Clinically relevant sources of variation (e.g. 100K genomes, gnomAD and ClinVarterm SNPs) are mapped to protein features and variants using a pre-calculated mappingterm of the genomic coordinatesterm for the amino acids at the beginning and end of each exon and the conversion of UniProt position annotations to their genomic coordinatesterm ( 30 ). Functional positional annotations from the UniProt human reference proteometerm are now being mapped to the corresponding genomic coordinatesterm on the GRCh38 version of the human genometerm for each release of UniProt. These mappings are also available as BED files or as part of a UniProt genomic track hub and can be downloaded from the UniProt FTP site ( www.uniprot.org/downloads ). Aligning variants to protein featuresterm, such as functional domainsterm and active sites, ligand binding sites and PTMs in the UniProt record, can provide mechanistic insightsterm into how specific variantsterm can lead to diseaseterm or resistance to a drugterm or to a pathogen. UniProt additionally integrates and visualizes unique and non-unique peptides identified by mass spectrometry proteomic dataterm deposited through the ProteomeXchange Consortiumterm ( 31 ) (e.g. PeptideAtlas ( 32 ), MassIVE ( 33 ) and jPOST ( 34 )) and other large-scale initiatives (CPTAC ( 35 ), ProteomicsDB ( 36 ), MaxQB ( 37 ), ETD and CTDP ( 38 )). Peptide identificationsterm can be taken as evidence that a protein has been validatedterm (PE = 1) using a variation of the HPP guidelines (in brief, at least two unique peptides of seven amino acids or more or, for proteins where this cannot be achieved due to sequence constraints, one unique peptide of ten amino acids or more has been mapped to a protein). UniProt also provides the new format PEFFterm (PSI Extended FASTA Formatterm) proposed by the HUPO-PSI (Human Proteometerm Organization-Proteomics Standardterm Initiative) for sequence databases ( 39 ) to be used by sequence search engines and other associated tools (e.g. spectral libraries search tools). The UniProt PEFFterm format currently represents variation and sequence data and is available through the Proteins APIterm service ( https://www.ebi.ac.uk/proteins/api/doc/ ) for 31 species. Community curation in UniProt UniProt users have always actively engaged with us and provide important feedback to the resource. The significant number of requests we receive through the Help Desk for articles and annotationsterm to be added to protein entries prompted the development of the ‘Community submissionterm’ page, where researchers are able to add articles that they deem relevant to an entry and provide optional basic annotation by selecting the topics relevant to each paper from a controlled listterm and/or adding short statements about protein name, function, and diseaseterm, as described in the publication. Contributors are asked to supply their ORCIDterm ( https://orcidterm.org/ ), a researcher personal IDterm, which is used to both validate that the submission is genuine and to give credit to the submitter for their work (Figure 5 ). This enables us to leverage the scientific community as a resource for enhancing our curated content, emulating a model already adopted by a number of model organism databasesterm, such as WormBase ( 40 ), PomBase ( 41 ) and FlyBase ( 42 ). Figure 5. Community contributionterm. (i) Use ‘Add a publication’ functionality (red box) in the UniProtKB entry. (ii) Partial snapshot of the submission form, a sample available here: https://community.uniprot.org/bbsub/sampleform.html . (iii) After submission and review the publication and information are displayed in the relevant UniProtKB entry with attribution to submitter (red box) in a future public release. Submissions are minimally checked by an experienced curatorterm before being added to the Publications section of the record. A tracker toolterm has also been developed ( https://community.uniprot.org/bbsub/bbsubinfo.html ) to enable users to access and search this wealth of additional bibliography. Researchers are encouraged to add relevant publications to entries of interest to them. Since we started the pilot in release 2019_08 we have seen a continuing increase in user submissions. As of release 2020_04 there have been 674 submissions relating to 424 publications and 557 entries, from 149 unique users ( https://community.uniprot.org/bbsub/STATS.html ). Website updates Users are at the center of the UniProt website design and development process. We follow a user-centered design processterm, conducting regular workshops, user testing, surveys and user research activities involving many users worldwide with varied research backgrounds and use cases. We are currently redesigning the website to take advantage of new technologies and paradigms in web development and welcome the UniProt user community to participate in user feedback and testing activities by contacting us at help@uniprot.org and contributing your user stories and workflows. As part of the redesign process, we have redeveloped both the UniProt interaction viewerterm (Figure 6A ) and ProtVista viewerterm, which graphically represents protein sequence features, such as domains, sites, post-translational modifications and variants, and structures. These are now part of the ‘Nightingaleterm’ visualization web component library ( https://ebi-webcomponents.github.io/nightingaleterm/#/ ) and are publicly available as lightweight, flexible, and modular componentsterm that can be more easily extended with new features, modified and implemented by users in their own resources (Figure 6B ). The ProtVista viewerterm has already been implemented by the Open Targets ( 43 ) and the Pharos ( 44 ) databases of unstudied and understudied drug targets amongst others. Figure 6. ( A ) The UniProtKB interaction viewer as seen in entry UniProtKB: Q9NSA3 , the beta-catenin-interacting protein 1. ( B ) The interaction viewer reusable web component in the Nightingaleterm library. DATA AVAILABILITY Due to the ever-increasing number of sequence records UniProt is processing with every release cycle, as of release 2020_01 (26 February 2020), UniProt releases are now published every eight weeksterm. This gives our production team the time required to complete data import, proteome redundancy removalterm, data checking, integration of external data and automatic annotationterm of unreviewed records prior to starting the release process. In addition to providing customizable views and downloads in a range of formats via the website, and file sets at the FTP site ( www.uniprot.org/downloads ), UniProt supplies users with a number of different options for computational access to the data ( www.uniprot.org/help/programmatic_access ). These include the website RESTful Application Programming Interfaceterm (API), stable URLsterm that can be bookmarked, linked, and reused, the Proteins extended REST API providing genomic coordinatesterm of UniProtKB sequences and annotations imported and mapped from large-scale data imports ( 29 ), and the SPARQL APIterm that allows users to perform complex queries across all UniProt data and also other resources that provide a SPARQL endpointterm, such as DisGeNET ( 45 ), Bgee ( 46 ), or Wikidata ( 47 ). The Proteins APIterm has recently been extended to serve the HUPO Proteomics Standardsterm Initiative Extended FASTA Format (PEFFterm) for the proteomics community which enables more metadata, such as detail of amino-acid variants in the FASTA file header section ( 39 ). Outreach and training The training of our users to make the best use of our data continues to be of critical importance to the UniProt Consortium. In order to access our global user base, we are increasingly moving away from classroom style training towards use of distance learning techniques. We regularly run webinars, which range from basic introductions to the use of the web pages aimed at new users and students, to more specialist sessions on, for example, use of the API or 3D structural representation in UniProtKB. The webinars are recorded and are subsequently made available online ( https://www.ebi.ac.uk/training/online/ ) where they are supported by related online training materials and YouTube videos ( https://www.youtube.com/user/uniprotvideos/ ). In order to reach new communities, the webinars are widely advertised through social media forums such as Twitter (@uniprot) and Facebook ( https://www.facebook.com/uniprot.org/ ) as well as established mailing lists. CONCLUSIONS UniProt continues to play its pivotal role in the fields of biology and biomedicine, collecting, standardizing and organizing knowledge of proteins and their functions to create a reference framework for multiscale biomedical data integration and analysisterm. Organisms are being routinely sequenced at the whole genome level, and eukaryotic, prokaryotic, and metagenomic sequencing projects are all contributing to the increased diversity of sequence data in the UniProt databases. It is of increasing importance that our automatic annotation pipelinesterm continue to develop in parallel to ensure that these unreviewed genomes, the vast majority of which are not being experimentally studied at the protein level, are richly and comprehensively annotated with functional informationterm. Expert curationterm of those proteins biochemically characterized remains a key focus of our activities, to both inform on these well-studied entities and also to act as template entriesterm for information transfer to proteins in related speciesterm. As the complexity and depth of our value-added data increases, we are exploring new ways to present the data to users and will continue to serve the community with new and improved website access designed to improve and enhance the user experience and upgraded programmatic accessterm, with ease of use always a priority. We greatly value the feedback and annotation updates from our user community. Please send your feedback and suggestions to the e-mail address help@uniprot.org or via the contact link on the UniProt website. ACKNOWLEDGEMENTS The UniProt publication has been prepared by Alex BatemanpersName, Maria-Jesus MartinpersName, Sandra OrchardpersName, Michele MagranepersName, Rahat AgivetovapersName, Shadab AhmadpersName, Emanuele AlpipersName, Emily H. Bowler-BarnettpersName, Ramona BrittopersName, Borisas BursteinaspersName, Hema Bye-A-JeepersName, Ray CoetzeepersName, Austra CukurapersName, Alan Da SilvapersName, Paul DennypersName, Tunca Dogan, ThankGod EbenezerpersName, Jun FanpersName, Leyla Garcia CastropersName, Penelope GarmiripersName, George GeorghioupersName, Leonardo GonzalespersName, Emma Hatton-EllispersName, Abdulrahman HusseinpersName, Alexandr IgnatchenkopersName, Giuseppe InsanapersName, Rizwan IshtiaqpersName, Petteri JokinenpersName, Vishal JoshipersName, Dushyanth JyothipersName, Antonia LockpersName, Rodrigo Lopez, Aurelien LucianipersName, Jie LuopersName, Yvonne LussipersName, Alistair MacDougallpersName, Fabio MadeirapersName, Mahdi MahmoudypersName, Manuela MenchipersName, Alok MishrapersName, Katie MoulangpersName, Andrew Nightingaleterm, Carla Susana OliveirapersName, Sangya PundirpersName, GuoyingpersName Qi, Shriya RajpersName, Daniel RicepersName, Milagros Rodriguez Lopez, Rabie SaidipersName, Joseph SampsonpersName, Tony SawfordpersName, Elena SperettapersName, Edward TurnerpersName, Nidhi TyagipersName, Preethi VasudevpersName, Vladimir VolynkinpersName, Kate WarnerpersName, Xavier WatkinspersName, Rossana ZarupersName, and Hermann ZellnerpersName at the EMBL- European Bioinformatics Institute; Alan BridgepersName, Sylvain PouxpersName, Nicole RedaschipersName, Lucila AimopersName, Ghislaine Argoud-PuypersName, Andrea AuchinclosspersName, Kristian AxelsenpersName, Parit BansalpersName, Delphine BaratinpersName, Marie-Claude BlatterpersName, Jerven BollemanpersName, Emmanuel BoutetpersName, Lionel BreuzapersName, Cristina Casals-CasaspersName, Edouard de CastropersName, Kamal Chikh EchioukhpersName, Elisabeth CoudertpersName, Beatrice CuchepersName, Mikael DochepersName, Dolnide DornevilpersName, Anne EstreicherpersName, Maria Livia FamigliettipersName, Marc FeuermannpersName, Elisabeth GasteigerpersName, Sebastien Gehant, Vivienne GerritsenpersName, Arnaud GospersName, Nadine Gruaz-GumowskipersName, Ursula HinzpersName, Chantal HulopersName, Nevila Hyka-NouspikelpersName, Florence JungopersName, Guillaume KellerpersName, Arnaud KerhornoupersName, Vicente LarapersName, Philippe Le MercierpersName, Damien LieberherrpersName, Thierry LombardotpersName, Xavier MartinpersName, Patrick MassonpersName, Anne MorgatpersName, Teresa Batista NetopersName, Salvo PaesanopersName, Ivo PedruzzipersName, Sandrine PilboutpersName, Lucille PourcelpersName, Monica PozzatopersName, Manuela PruesspersName, Catherine RivoirepersName, Christian SigristpersName, Karin SonessonpersName, Andre StutzpersName, Shyamala SundarampersName, Michael TognollipersName, and Laure VerbreguepersName at the SIB Swiss Institute of Bioinformatics; CathypersName H. Wu, Cecilia N. ArighipersName, Leslie ArminskipersName, Chuming ChenpersName, Yongxing ChenpersName, John S. GaravellipersName, Hongzhan HuangpersName, Kati LaihopersName, Peter McGarveypersName, Darren A. NatalepersName, Karen RosspersName, C. R. VinayakapersName, Qinghua WangpersName, Yuqi WangpersName, Lai-Su YehpersName, and Jian ZhangpersName at the Protein Information Resource. The UniProt consortium also acknowledges the contributions of Patrick RuchpersName and Douglas TeodoropersName from the SIB Swiss Institute of Bioinformatics who developed the UPCLASS publication classification tool and the ChEBI team at EMBL-EBI for their support in small molecule annotation. Contributor Information The UniProt Consortium, European Molecular Biology Laboratory, European Bioinformatics Institute (EMBL-EBI), Wellcome Genome Campus, Hinxton CB10 1SD, UK; Protein Information Resource, Georgetown University Medical Center, 3300 Whitehaven Street NW, Suite 1200, Washington, DC 20007, USA; Protein Information Resource, University of Delaware, Ammon-Pinizzotto Biopharmaceutical Innovation Building, Suite 147, 590 Avenue 1743, Newark, DE 19713, USA; SIB Swiss Institute of Bioinformatics, Centre Medical Universitaire, 1 rue Michel Servet, CH-1211 Geneva 4, Switzerland. The UniProt Consortium: Alex BatemanpersName , Maria-Jesus MartinpersName , Sandra OrchardpersName , Michele MagranepersName , Rahat AgivetovapersName , Shadab AhmadpersName , Emanuele AlpipersName , Emily H Bowler-BarnettpersName , Ramona BrittopersName , Borisas BursteinaspersName , Hema Bye-A-JeepersName , Ray CoetzeepersName , Austra CukurapersName , Alan Da SilvapersName , Paul DennypersName , Tunca Dogan , ThankGod EbenezerpersName , Jun FanpersName , Leyla Garcia CastropersName , Penelope GarmiripersName , George GeorghioupersName , Leonardo GonzalespersName , Emma Hatton-EllispersName , Abdulrahman HusseinpersName , Alexandr IgnatchenkopersName , Giuseppe InsanapersName , Rizwan IshtiaqpersName , Petteri JokinenpersName , Vishal JoshipersName , Dushyanth JyothipersName , Antonia LockpersName , Rodrigo Lopez , Aurelien LucianipersName , Jie LuopersName , Yvonne LussipersName , Alistair MacDougallpersName , Fabio MadeirapersName , Mahdi MahmoudypersName , Manuela MenchipersName , Alok MishrapersName , Katie MoulangpersName , Andrew Nightingaleterm , Carla Susana OliveirapersName , Sangya PundirpersName , GuoyingpersName Qi , Shriya RajpersName , Daniel RicepersName , Milagros Rodriguez Lopez , Rabie SaidipersName , Joseph SampsonpersName , Tony SawfordpersName , Elena SperettapersName , Edward TurnerpersName , Nidhi TyagipersName , Preethi VasudevpersName , Vladimir VolynkinpersName , Kate WarnerpersName , Xavier WatkinspersName , Rossana ZarupersName , Hermann ZellnerpersName , Alan BridgepersName , Sylvain PouxpersName , Nicole RedaschipersName , Lucila AimopersName , Ghislaine Argoud-PuypersName , Andrea AuchinclosspersName , Kristian AxelsenpersName , Parit BansalpersName , Delphine BaratinpersName , Marie-Claude BlatterpersName , Jerven BollemanpersName , Emmanuel BoutetpersName , Lionel BreuzapersName , Cristina Casals-CasaspersName , Edouard de CastropersName , Kamal Chikh EchioukhpersName , Elisabeth CoudertpersName , Beatrice CuchepersName , Mikael DochepersName , Dolnide DornevilpersName , Anne EstreicherpersName , Maria Livia FamigliettipersName , Marc FeuermannpersName , Elisabeth GasteigerpersName , Sebastien Gehant , Vivienne GerritsenpersName , Arnaud GospersName , Nadine Gruaz-GumowskipersName , Ursula HinzpersName , Chantal HulopersName , Nevila Hyka-NouspikelpersName , Florence JungopersName , Guillaume KellerpersName , Arnaud KerhornoupersName , Vicente LarapersName , Philippe Le MercierpersName , Damien LieberherrpersName , Thierry LombardotpersName , Xavier MartinpersName , Patrick MassonpersName , Anne MorgatpersName , Teresa Batista NetopersName , Salvo PaesanopersName , Ivo PedruzzipersName , Sandrine PilboutpersName , Lucille PourcelpersName , Monica PozzatopersName , Manuela PruesspersName , Catherine RivoirepersName , Christian SigristpersName , Karin SonessonpersName , Andre StutzpersName , Shyamala SundarampersName , Michael TognollipersName , Laure VerbreguepersName , CathypersName H Wu , Cecilia N ArighipersName , Leslie ArminskipersName , Chuming ChenpersName , Yongxing ChenpersName , John S GaravellipersName , Hongzhan HuangpersName , Kati LaihopersName , Peter McGarveypersName , Darren A NatalepersName , Karen RosspersName , C R VinayakapersName , Qinghua WangpersName , Yuqi WangpersName , Lai-Su YehpersName , Jian ZhangpersName , Patrick RuchpersName , and Douglas TeodoropersName FUNDING National Eye Institute, National Human Genome Research Institute, National Heart, Lung, and Blood Institute, National Institute of Allergy and Infectious Diseasesterm, National Institute of Diabetes and Digestive and Kidney Diseasesterm, National Institute of General Medical Sciences, National Cancer Institute, National Institute On Aging, and National Institute of Mental Health of the National Institutes of Health [U24HG007822]; National Human Genome Research Institute [U41HG002273]; National Institute of General Medical Sciences [R01GM080646, P20GM103446] (the content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health); Biotechnology and Biological Sciences Research Council [BB/T010541/1]; British Heart Foundation [RG/13/5/30112]; Open Targets; Swiss Federal Government through the State Secretariat for Education, Research and Innovation SERI; European Molecular Biology Laboratory core funds. Funding for open access charge: National Institutes of Health [U24HG007822]. Conflict of interest statement . None declared.