Using UniProtKB to navigate large and complex structures

With advances in structural biology, protein structures are becoming larger and more complex than ever. How do we navigate these complex structures?

PDB:8F2U captures the human COMMD–CCDC22–CCDC93 (CCC) complex, which is part of the Commander complex that plays a vital role in sorting transmembrane cargo for endosomal recycling. The CCC complex has 12 different protein components, but how do we identify the functional role of each protein?

A handy tool is UniProt's advanced search function, where you can find all the protein components using the PDB entry ID (xref:pdb-8F2U). This directs you to the individual entry pages, where you can learn about the biological role of each protein.

You can also use the ‘Customize columns’ option to include 3D structures as an additional column. This provides a glimpse of the available structures for each protein.

Which structure is the ‘best’ structure?

There is often more than one experimentally determined structure available for a protein and they might provide different information. In the UniProt entry page, you can go to the Structure section for a list of all the available structures. Let’s take a look at COMMD7 (AC:Q86VX2).

4 structures are available for this protein and they were solved using different methods with varying resolutions. X-ray crystallography often provides high-resolution structures for smaller proteins, while cryo-EM can capture the conformation of larger and more dynamic protein complexes. Understanding how the structures were solved can help us select the most informative structure for our study.

Another useful piece of information is the ‘POSITIONS’ column. The positions here indicate the construct used to determine the structure and they may not cover the full-length protein. In this example, if you are interested in the N-terminal 130 residues of COMMD7, they are found in PDB:8F2R, 8F2U, 8P0W, but not in PDB:8ESD. In some cases, even if a full-length construct is used, some regions may not be observable in the structure. This is usually due to technical limitations in resolving highly dynamic regions in a protein.

If you are interested in the conformation of the full-length protein, you can refer to the AlphaFold prediction. UniProt provides an AlphaFold model for most protein entries, predicted based on the canonical sequence. These full-length models can provide insights for regions that are absent in experimentally determined structures.

Where is my residue of interest?

A high-resolution protein structure can provide residue-level information about a protein. For example, post-translational modification (PTM) analysis by the PTMeXchange project showed that residue Lys90 in human COMMD7 is ubiquitinated. But where is it in the structures?

UniProt and PDB entries provide a consistent residue-level mapping through the SIFTS project. This means for all the PDB structures mapped to the same UniProt entry, you can expect to see the same Lys90 in the structures covering this region of the protein.

Using the UniProt Feature viewer, you can identify where Lys90 is.

Clicking on Lys90 will highlight this residue across all the feature viewer tracks. It will also zoom in and highlight Lys90 in the structure viewer. But don't forget to select a structure that covers residue 90! (Hint: not PDB:8ESD)

How to use this information for further studies?

If you want to study these structures in detail, you can directly download them from the UniProt entry page.

PyMOL and UCSF Chimera are among the most common molecular visualisation tools in bioinformatics and computational chemistry. You can use them for more complex analysis, such as measuring bond lengths, docking small molecules and simulating conformational changes.

We can use the information from UniProt to get us started. For example, knowing that chain G in PDB:8F2U is COMMD7, you can highlight COMMD7 in the structure of the 12-subunit CCC complex. You can also identify Lys90 and check if it is accessible for ubiquitination in the complex.

Try these commands in PyMOL:

fetch 8f2u

select COMMD7, 8f2u and chain G

util.cbay COMMD7

show sticks, (COMMD7 and resi 90 and not name N+C+O)

label n. CA and i. 90 and COMMD7, '%s%s' % (resn, resi)

Have fun navigating the world of protein structures!

Learn more about structural annotations in UniProt: https://www.uniprot.org/help/structure_section

Learn more about multiple structures for the same protein:
https://www.uniprot.org/help/multiple_pdb_xrefs

Learn more about sequence coverage in structures:

https://www.uniprot.org/help/structure_subseq

UniProt - the ultimate colleague on your biological research team!

 

How many members do you have on your team and have you ever considered UniProt as one of them? 

UniProt is a suite of open access protein databases, accessed by 9 million unique visitors a year, but how much money does it save you?

Our contribution to the scientific community and wider economy has now been analysed in a case study by CSIL, as part of the EU-funded project PathOS. This study investigated the cost-benefit of open data resources, and an analysis of UniProt’s impact between the years of 2017 and 2023 has been published.

What are the main costs of maintaining UniProt?
Across our three consortia sites, EMBL-EBI, SIB, PIR, we have office space, equipment, consumables and publication costs. But our main outgoings (70%) are salaries for our teams of expert biocurators and software developers. Cost to a user is measured in time spent providing valuable voluntary knowledge contributions. 49% of our users visit weekly and 26% daily.

What are the benefits to using UniProt?
The immediate benefits for users are significant time savings. UniProt integrates data from over 180 cross-reference databases, combines it with expert manually curated sequence and protein data, then puts it all in one place, in a standardized format. This saves users time because they don’t have to navigate between multiple resources or download data in different formats. Additionally users do not have to sift through numerous scientific publications to understand the current cutting edge research available for a given protein. It also helps streamline research and analysis pipelines by providing centralized ready-to-use data. 

Long-term benefits are measured in publications and patents that mention UniProt, indicating the wider impact on scientific knowledge advancement and technological development. Over a period of 7 years over 15,200 publications and over 183,000 patents cited or referenced UniProt. Many of these patents go on to be referenced by subsequent patents spanning a number of biotechnology and health innovation fields. Economic benefits have also been seen in the number of start-up companies that rely on UniProt data for their business model.

Demonstrating the value of UniProt

• Each user gains a net benefit from UniProt of up to €5,475 and saves 219 hours a year.
• Users say that our main strength ‘is the ability to integrate protein sequences identified in the literature with an extensive body of functional information’. This is facilitated by incredibly passionate teams of expert biocurators and software developers that ensure data is curated into the database reliably, and presented in a consistent and easily accessible manner.
• Users agree ‘that there is no alternative offering the same breadth of knowledge, quality and level of integration as UniProt’. 
• Overall users say: ‘UniProt plays a crucial role in accelerating scientific research and innovation in various forms, thereby facilitating the creation of new knowledge.’
• In total UniProt provides a benefit of between €373-565 million per year to its community of scientific users.

Learn more about the details of how UniProt supports research and innovation by reading the full report: Measuring the value and impact of open science

Links to associated articles

ELIXIR - https://elixir-europe.org/news/UniProt-CBA

SIB -  https://www.sib.swiss/news/uniprot-user-benefits-up-to-39-times-higher-than-operational-costs

PathOS - https://pathos-project.eu/open-science-value-costs-and-benefits-for-whom-how-to-support-informed-investment-decisions

The UniProt team would like to take the opportunity to thank our funders, collaborators and users for their time, support and contributions to the database. UniProt is a key part of the  EMBL-EBI, SIB Swiss Institute of Bioinformatics and the Protein Information Resource (PIR)  and ELIXIR

infrastructure. Our main funders are EMBL, The State Secretariat for Education, Research and Innovation SERI (Switzerland), and NIH (USA). We are one of ELIXIR’s Core Data Resources, and two of our three partners, EMBL-EBI and SIB, are ELIXIR Nodes. The findings of this study will support efforts to advocate for long-term funding for critical biodata resources.

Capturing the Diversity of Life - Reorganizing the Protein Space in UniProtKB

 

Advances in genome sequencing technology means that large-scale efforts such as the Earth Biogenome project and the Darwin Tree of Life are aiming to produce high-quality reference genomes for individual species, to capture the biodiversity of our planet. Each of these genomes will translate to complete set of proteins for that species. To enable UniProt to present this wealth of data to our users, we are making significant improvements to our data content.

 

Upcoming changes in our selection of Reference Proteomes

 

In release 2025_04 (currently scheduled 8th October 2025), we will deploy a new Reference Proteome selection pipeline to improve the representation of species biodiversity in the UniProt Knowledgebase (UniProtKB). From release 2026_01 onwards, we will restrict the protein space in UniProtKB to those sequences which are part of a Reference Proteome in addition to the expert reviewed UniProtKB/Swiss-Prot section, and also unreviewed entries associated with experimental Gene Ontology annotations or additional biologically important data such as a 3D structure.

 

What is a Reference Proteome?

A proteome is the set of all translated proteins from a genome assembly. For each species, we generally use an automatic pipeline to select one proteome as the Reference Proteome - the proteome that we believe is the best representative of the proteins encoded by that species (for example, the Reference Proteome for Drosophila simulans is UP000000304 as it provides the best coverage of the protein space for the species). In addition, proteomes of well-studied model organisms and other proteomes of interest for biomedical and biotechnological research may be selected as Reference Proteomes by UniProt curators.

 

Is every sequenced proteome currently in UniProtKB?

No. While we currently include many proteomes that are not Reference Proteomes in UniProtKB, we already exclude many others, for example those which are of poor quality or where the species is already over-represented in UniProtKB.

 

Why are we reorganizing the protein space in UniProtKB?

Submissions of genomes to hubs such as the International Nucleotide Sequence Database Collaboration (INSDC) have grown due to the rapid increase in sequencing capabilities, resulting in a large influx of proteomes into UniProtKB. Our new pipeline will provide a much better representation of the biodiversity of life and coverage of the sequence space and improve user experience when searching and selecting the best proteome for their research work. It will also allow us to improve our functional annotation of the proteins in these proteomes in order to provide an enhanced understanding of these species.

 

Figure 1 - Growth of UniProtKB entries throughout the years

 

What is going to happen over the next few months?

In release 2025_04, we will deploy the new pipeline that will improve our selection of Reference Proteomes for cellular organisms (more details below), and we will align our viral Reference Proteomes to the set of exemplar genomes from the International Committee on Taxonomy of Viruses (ICTV). Additionally, we will start the process of removing proteins from non-Reference Proteomes from UniProtKB. In the first stage (2025_04), this includes proteins from taxonomically unclassified organisms. In release 2026_01 we will remove the remainder of proteins from non-Reference Proteomes from UniProtKB.

 

 

How does the new pipeline work?

The new pipeline to select Reference Proteomes has been designed to select the proteome that best represents the protein space of a species using a clustering system based on MMseqs2 [1] to select one or a few, proteomes for each species. It only analyzes high quality proteomes from species with a recognized taxonomy and a formal scientific name. Problematic proteomes, such as proteomes from contaminated genomes, for example, are excluded from the analysis.

 

MMseqs2 is used to cluster first proteins, then proteomes:

1. Protein clusters: firstly, it clusters similar proteins from different proteomes of the same species. The more proteins a proteome has in different clusters, the more likely it is to be selected as a reference.

2. Proteome clusters: secondly, it calculates how similar the proteomes of the same species are, based on the number of protein clusters shared between any two proteomes. Proteomes are clustered together if they share 50% or more protein clusters.

 

Figure 2 - Diagram of the Reference Proteome selection pipeline

 

This new pipeline will select at least one Reference Proteome for every sequenced species for inclusion in UniProtKB, ensuring a broad representation of the Tree of Life. It may select more than one  Reference Proteome for species with a high genomic diversity. To keep the set of Reference Proteomes stable, but not static, between UniProt releases the pipeline preferably selects the most complete proteome (as determined by BUSCO [2]) and will only replace an existing Reference Proteome when a new one is published which is of significantly higher quality.

 

How big a change will this make to UniProtKB?

The changes in UniProt, between release 2025_04 and 2026_01, will result in a change in the content and size of the database. The number of Reference Proteomes will increase by 36% (reflecting a 34% increase on species covered), while the number of proteins in UniProtKB can be decreased by 43%

 

What do I do if the proteome I am working on is no longer in UniProtKB?

All proteomes that are not selected as Reference Proteomes by the new pipeline will be available through UniParc. When you search for such proteomes in the Proteomes portal, you will be directed to UniParc to access your protein set. The UniParc FASTA header format for proteomes will be improved to show the protein and gene names, and database identifiers, of the underlying genome records from EMBL, Ensembl or RefSeq.

 

If the annotation provided by UniProtKB is particularly important to your work, or your organism is actively worked on by a research community, but has not been selected as a Reference Proteome, please Contact us and we will consider promoting it to Reference Proteome status. The list of proteomes that will be deprecated by release 2026-01, and also the proteins associated with them, is available on our FTPsite. The same link will allow you to access an additional document which lists all the proteins we will be retaining in UniProtKB.

 

 

[1] MMseqs2 enables sensitive protein sequence searching for the analysis of massive data sets | Nature Biotechnology

[2] BUSCO Update: Novel and Streamlined Workflows along with Broader and Deeper Phylogenetic Coverage for Scoring of Eukaryotic, Prokaryotic, and Viral Genomes | Molecular Biology and Evolution | Oxford Academic

 

 

 

Proteome help page: https://www.uniprot.org/help/proteome

MMseqs2 GitHub: https://github.com/soedinglab/MMseqs2

 

 Rich Epitope Information Comes to UniProt

Mammalian immune responses are mediated by interactions between antigens and

immune system components such as antibodies, B cells, and T cells. However,

antibodies and immune cells do not bind to entire antigens, which are usually

proteins or large polysaccharides; instead, they recognize one or more small

regions within the antigen called epitopes. Characterizing epitopes gives us

insight into infectious diseases, autoimmune diseases, and cancer,

and leads to therapeutic innovations such as the development of more effective

vaccines.

UniProt curators have traditionally included information about protein epitopes

from the literature as part of the process of manually annotating protein entries.

However, epitope information in UniProt recently got a big boost from a

collaboration with the Immune Epitope Database (IEDB). The IEDB is a

freely available, manually curated resource that catalogs experimental data

on antibody and T cell epitopes in humans and other animal species in

the context of a variety of diseases and conditions.

Thanks to the UniProt-IEDB collaboration, epitopes curated by the IEDB can be viewed

in a track in the UniProt Feature Viewer with links back to the IEDB. In addition,

publications with epitope information identified by the IEDB are now accessible on UniProt

Publications pages, and IEDB epitopes are searchable using the Proteins API.

The collaboration has enhanced UniProt with information about more than 700,000

naturally occurring, linear peptide epitopes in over 57,000 proteins, citing over 7,000 papers that describe their experimental characterization.

For example, consider the protein O-phosphoseryl-tRNA(Sec) selenium transferase

(SEPSECS; UniProt ID: Q9HD40). SEPSECS (aka SLA/LP autoantigen), which

normally functions as an enzyme in selenoprotein biosynthesis, is an autoantigen in

autoimmune hepatitis (AIH), a chronic inflammatory disease of the liver. Patients

with AIH have circulating antibodies against SEPSECS as well as lymphocyte

infiltrations in the liver. The UniProt entry page identifies the region from amino acids

474-493 as an SLA/LP epitope based on UniProt curation of a publication

characterizing autoantibodies in AIH (PMID:11826415; top panel of figure). The

Epitopes track of the Feature Viewer (middle panel of figure) displays the epitopes of

SEPSECS that have been curated by the IEDB, aligned with their positions in the

protein sequence. Clicking on an epitope brings up a box with the epitope sequence,

the experiments in which the epitope was studied, and a link to the epitope’s page in IEDB.

Finally, the Publications page (bottom panel of figure) lists a paper with epitope

information that was cited by IEDB (PMID:18773898). According to the accompanying

annotation, SEPSECS is a target of T cells in patients with AIH. A review of the abstract

reveals that the paper describes the identification of multiple epitopes in SEPSECS that

are recognized by autoreactive CD4+ T cells in AIH.

The inclusion of epitope and related immune information from IEDB in UniProt is an

exciting development that will hopefully prove valuable to immunologists and others

interested in the role of the immune system in health and disease.

 

#UsingUniProt - DisCanVis interpreting genomic variation data

Norber Deutsch, Mátyás Pajkos, Gábor Erdős and Zsuzsanna Dosztányi

Department of Biochemistry, Institute of Biology, ELTE Eötvös Loránd University, Budapest, Hungary.

In recent years a wealth of information has become available about genetic variations that underlie various diseases, especially cancer. However, the interpretation of these variations is far from trivial. Many of the observed mutations do not directly contribute to the disease development or have no known function associated with it. Protein level information often proves to be essential to guide us towards biologically relevant cases.

Currently, genome and protein level information are available through distinct resources with very limited overlap. Databases, such as COSMIC or CbioPortal serve as entry points for accessing genetic variations (Tate et al. 2019; Gao et al. 2013). However, the central resource for protein level information is the UniProtKB database which provides a rich source of annotations about the structural and functional properties of proteins. In order to help the interpretation of genetic variations, researchers from the Dosztányi lab at the Eötvös Loránd University in Budapest developed a novel web-based visualization tool, called DisCanVis (http://discanvis.elte.hu), which can bring these two worlds together (Deutsch et al. 2023).

When we consider a genetic variation, one of the basic questions we can ask is whether the variation affects a known structure. This can be important to model the impact of the mutations and to find drug molecules that can compensate for the effect of mutations. However, we now know that a significant portion of the human genome encodes proteins whose native state cannot be characterized by a single well-defined structure (Dyson and Wright 2005). These so-called intrinsically disordered protein and protein regions (IDPs/IDRs) can only be represented by an ensemble of different conformations that cannot be accurately captured even by the recent AlphaFold method which achieved a significant breakthrough in predicting the structures of globular proteins (Jumper et al. 2021; Ruff and Pappu 2021)  .

IDRs carry out important functions in many regulatory signaling processes. The key to this is their dynamic nature which enables them to form interactions that can be quickly turned on or off depending on cellular cues. Such interactions are also critical for many proteins that are known to be involved in cancer, such as p53. However, the direct role of IDRs in cancer is much less well-understood.

One of the main focuses of DisCanVis is to help the study the role of IDRs in cancer and to help to answer questions such as: What is the role of disordered regions in cancer proteins? In which cases do cancer mutations specifically target intrinsically disordered regions? What kind of annotations can be found for the mutated regions?

DisCanVis is built over 18,000 human proteins that are shared between COSMIC and Uniprot databases. The server combines cancer and other disease variations with protein level functional and structural annotations collected through the UniProt database. A key element of collected features is related to protein disorder and includes annotations of experimentally verified disordered regions, state-of-the-art prediction methods to reliably assess protein disorder and the locations of known functional regions within IDPs, such as short linear motif sites, disordered binding regions, and post-translational modifications. Altogether, more than 30 different features are collected and projected along the sequence for the visualization. Entries can be searched by UniProt accession number, protein or gene names.

Figure 1 (The concept behind DisCanVis.)

We present the usage of DisCanVis through the example of β-catenin (P35222). This protein is a key regulator of cell growth and survival through the Wnt signaling pathway. β-catenin is frequently mutated in various types of cancer, including colorectal, liver, breast, and lung cancers. Mutations are enriched in a short region within a disordered segment at the N-terminal part of the protein. The region corresponds to a β-TrCP binding motif, which, under normal conditions, is recognized by the β-TrCP E3 ligase which regulates the degradation of beta-catenin. The mutations of the binding motif interfere with the proper degradation of beta-catenin, resulting in its pathological accumulation in the cell, which can lead to the activation of genes that drive uncontrolled cell growth and tumor formation (Bugter, Fenderico, and Maurice 2021).

Figure 2 (DisCanVis visualization for β-catenin)

The visualization can be separated into three sections: The first section is the header which shows the full-length protein, giving an overview of the mutations and structural state of the protein along the sequence. Below that the genome level annotations are presented indicating mutations collected from various cancer samples and other disease mutations. This is followed by the protein level annotations, including known structures, domains, and the disordered specific annotations. Annotated functional sites are also indicated, including known short linear motifs. The yellow box indicates mutation hotspots which in this case highlights the β-TrCP binding motif and phosphorylation sites which are critical for cancer development.

In addition to inspecting individual proteins, users can also carry out analyses by focusing on specific subsets of proteins through pre-compiled tables. These can be sorted and filtered enabling users to collect examples with existing annotations of protein disorder and associated functions, or discover currently uncharacterized examples with likely disease relevance. For example, it is possible to browse tables of known cancer drivers, experimentally verified disordered proteins and known linear motif sites or to explore proteins with a given Gene Ontology term. Users can also find proteins with a given type of short linear motif with the largest number of disease mutations, or to find regions that are enriched in mutations in a yet unclassified cancer driver.

 

DisCanVis combines the wealth of information on genetic variations with the highly valuable annotations largely expertly curatedmanually collected in the UniProt database. Through this, hopefully it will prove to be a valuable tool for advancing our understanding of intrinsically disordered proteins and to gain important insights into their roles in cancer.

Further information about the usage and functionalities can be found either in the web server or in the publication which can be found here:

Protein Science Volume32, Issue1
https://doi.org/10.1002/pro.4522

REFERENCES

Bugter, Jeroen M., Nicola Fenderico, and Madelon M. Maurice. 2021. “Mutations and Mechanisms of WNT Pathway Tumour Suppressors in Cancer.” Nature Reviews. Cancer 21 (1): 5–21.

Deutsch, Norbert, Mátyás Pajkos, Gábor Erdős, and Zsuzsanna Dosztányi. 2023. “DisCanVis: Visualizing Integrated Structural and Functional Annotations to Better Understand the Effect of Cancer Mutations Located within Disordered Proteins.” Protein Science: A Publication of the Protein Society 32 (1): e4522.

Dyson, H. Jane, and Peter E. Wright. 2005. “Intrinsically Unstructured Proteins and Their Functions.” Nature Reviews. Molecular Cell Biology 6 (3): 197–208.

Gao, Jianjiong, Bülent Arman Aksoy, Ugur Dogrusoz, Gideon Dresdner, Benjamin Gross, S. Onur Sumer, Yichao Sun, et al. 2013. “Integrative Analysis of Complex Cancer Genomics and Clinical Profiles Using the cBioPortal.” Science Signaling 6 (269): l1.

Jumper, John, Richard Evans, Alexander Pritzel, Tim Green, Michael Figurnov, Olaf Ronneberger, Kathryn Tunyasuvunakool, et al. 2021. “Highly Accurate Protein Structure Prediction with AlphaFold.” Nature 596 (7873): 583–89.

Ruff, Kiersten M., and Rohit V. Pappu. 2021. “AlphaFold and Implications for Intrinsically Disordered Proteins.” Journal of Molecular Biology 433 (20): 167208.

Tate, John G., Sally Bamford, Harry C. Jubb, Zbyslaw Sondka, David M. Beare, Nidhi Bindal, Harry Boutselakis, et al. 2019. “COSMIC: The Catalogue Of Somatic Mutations In Cancer.” Nucleic Acids Research 47 (D1): D941–47.