1. Sandra Orchard EMBL-EBI
This presentation introduces the protein database UniProt.
I’ll start with an introduction to UniProt giving a bit of background and describing what we’re trying to achieve.
Then I’ll go through the various sections of a UniProtKB\Swiss-Prot entry to show you what kind of biological information we capture and present.
I’ll explain our Automatic Annotation systems, this is the way in which we copy data from well studies proteins to those which haven’t been studied but when predict have a similar function.
[Drop this for patent talks - I’ll talk a little bit of Proteomics, what is a proteome and how to access/download a proteome from UniProt.]
The to finish off I’ll go through the functionality of the uniprot.org webiste.

2. Background of UniProt
Since 2002 a merger and collaboration of three databases:
Swiss-Prot & TrEMBL
PIR-PSD
Prior to 2002 there used to be 3 separate protein databases located around the globe each with its own data and rules for annotation. In 2002 these databases merged to provide and unified source with common annotation standards and datasets.
UniProt is funded mainly by….
Funded mainly by NIH (US) to be the highest quality, most thoroughly annotated protein sequence database

3. We Aim To Provide… A high quality protein sequence database
A non redundant protein database, with maximal coverage including splice isoforms, disease variant and PTMs. Sequence archiving essential.
Easy protein identification
Stable identifiers and consistent nomenclature/controlled vocabularies
Thorough protein annotation
Detailed information on protein function, biological processes, molecular interactions and pathways cross-referenced to external source

5. The Two Sides of UniProtKB
UniProtKB/TrEMBL
1 entry per nucleotide submission
UniProtKB/Swiss-Prot
1 entry per protein
This is a very important aspect of is that there is two sides to the databases. Basically, UniProt entries which have been looked at by a Curator and those which haven’t. Those which haven’t been manually reviewed by a Curator reside in TrEMBL. These are automatic translations of ENA the European Nucleotide Archive (formerly know as EMBL). The European Nucleotide Archive is stuctured such that each sequence is owned by the person which submitted it. Thus sequences can not be merged. Hence why when sequences are translated and incorporated in to UniProt they can be redundancy by that I mean more than one UniProt entry per protein. Also TrEMBL entries are unreviewed, thus they do not have annotation added from literature only minimal annotation from computational algorithms. By contrast, once a Curator has reviewed an entry and added information from all relevant literature the entry swaps to the Swiss-Prot side of the UniProt database. This is non-redundant as all entries for a protein are merged.
Indicators of which part of UniProt an entry belongs inlcude the colour of the starts and the ID....
Redundant, automatically annotated - unreviewed
Non-redundant, high-quality manual annotation - reviewed

6. Curation of a UniProt/SwissProt entry
Sequence
UniProt/TrEMBL
References
Sequence variants
Literature Annotations
Nomenclature
Ontologies
Sequence features
We’ll look at this in more detail after the break but briefly UniProt is the gold-standard resource for information on proteins. Every entry initially receives automatic annotation so its not just a bare sequence, there’s a team of Curators that undertake manual curation using the literature and sequence analysis. We also use in-house bioinformatics tools for protein classification and domain prediction. Data from other databases is imported and cross-referenced.
It comprises three different databases, but I haven’t shown all three here for the sake of simplicity. UniProtKB is the central database of protein sequences with accurate, consistent, and rich sequence and functional annotation. It comprises the manually annotated UniProtKB/Swiss-Prot section and the automatically annotated UniProtKB/TrEMBL section. The UniProt archive is an archive of all the protein sequences in the public domain, and the UniRef databases are a series of three databases that store sequences of 100%, 90% and 50% identity in the same records to speed up searching without losing information. UniProtKB contains more than 29 million cross references to over 100 other data resources; a few key ones are shown here.
UniProt/SwissProt

7. Searching UniProt – Simple Search
Text-based searching
Logical operators ‘&’ (and), ‘|’
Master headline

8. Searching UniProt – Advanced Search
Master headline

9. Each linked to the UniProt entry
Searching UniProt – Search Results
Each linked to the UniProt entry
Master headline

10. Searching UniProt – Search Results
Master headline

11. Searching UniProt – Search Results
Master headline

12. Searching UniProt – Blast Search
Master headline

13. As on slide
Just go through data types:
> Entry Name – if the entry has been reviewed then the first part should represent the gene name. The second part denotes the organism, usually it consists of the first 3 letters of the genus name and the first 2 letters of the species name, so for the Fruit fly Drosophila melanogaster all entries will be *****_DROME. Exceptions are HUMAN, MOUSE and RAT.
> Accession – an entry can have more than one accession if it is a Swiss-Prot.
> Entry history – (pretty self-explanatory) the Complete history takes you to page where any two versions from each UniProt release can be selected and compared.
> Entry status – whether its Reviewed (UniProtKB/Swiss-Prot) or Unreviewed (UniProtKB/TrEMBL).
> Annotation project – usually species specific.
> Disclaimer – only required for Human entry in case users try to self medicate base on the information we provide.

14. Protein names – There’s always a recommended name and if the protien is an enzyme the Enzyme Commision number is given which is a numerical classification scheme for enzymes, based on the chemical reactions they catalyze. Alternative names found in the literature are also added to aid finding the entry from a search.
Gene names – usually tries to represent the protein name but not always possible
Organism – with link to Complete proteome
Tax ID – Numerical UniProt taxon number and link to NCBI taxonomy browser
Taxonomic lineage.

15. Annotation comments FUNCTION PTM SUBCELLULAR LOCATION RNA EDITING
ALTERNATIVE PRODUCTS
MASS SPECTROMETRY
TISSUE SPECIFICITY
DOMAIN
DEVELOPMENTAL STAGE
POLYMORPHISM
INDUCTION
DISRUPTION PHENOTYPE
SIMILARITY
ALLERGEN
CATALYTIC ACTIVITY
DISEASE
COFACTOR
TOXIC DOSE
ENZYME REGULATION
BIOTECHNOLOGY
BIOPHYSICOCHEMICAL PROPERTIES
PHARMACEUTICAL
MISCELLANEOUS
PATHWAY
CAUTION
SUBUNIT
SEQUENCE CAUTION
INTERACTION
WEB RESOURCE
There’s a wide range of topics for us to captures as much information as possible from the literature, not every topic is present in every entry.
The green ones are very general, the blue ones are used a lot for enzymes and the pink ones are relevant to proteins involved in pathology/medicine.

16. Evidence tags to show source
Controlled vocabularies used whenever possible
Its in this section that all the information obtained from the literature is summarized and organized via specific fields.
Where possible we try to be consistent between protein with same function, this is aided by the use of controlled vocabularies.
Evidence tags are listed at the end of the section to show the source of the information.

17. As far as I am aware UniProt is unique in annotating features to areas of amino acid sequence.
As shown above we show regions of biological interest like those required for an interaction and involved in subcellular localization.
We also show smaller sites such as single amino acids that are modified post- translation.
Master headline

18. Automatic Annotation for UniProtKB/TrEMBL
Automatic annotation is a method of copying annotation to similar proteins, so those proteins that have not been studied gain some information.

19. UniProtKB/Swiss-Prot Manually annotated
UniProtKB/TrEMBL
Computationally
annotated
UniProtKB/Swiss-Prot
Manually
annotated
These graphs illustrate the need for automatic annotation.
The red dot puts the Swiss-Prot figure in the context of the TrEMBL graph
The message here is that there are a lot more entries TrEMBL than Swiss-Prot so we need to find a way to transfer annotation from Swiss-Prot entries to those that haven’t been reviewed in TrEMBL.
Web-pages for up to date Stats
UniProt/TrEMBL - UniProt/SwissProt -
Last updated 21/02/2012

20. InterPro
Master headline

21. Master headline

22. Automatic Annotation
UniProtKB employs two prediction programs which are referred to as UniRule and SAAS.
SAAS, Statistical Automatic Annotation System, generates a new set of decision-trees with every UniProtKB release using data-mining.
UniRule maintains a set of manually established and maintained annotation rules.
Automatic annotation is produced by two methods. One method just uses a computer program to generate rules, this is named SAAS, which stands for Statistical Automatic Annotation System. The other method is called UniRule and is a collection of rules created by Scientists to propagate a specific set of data based on a defined criteria. Both of these systems use Swiss-Prot and InterPro as training sets.
Swiss-Prot
InterPro

23. Help / Feedback help@uniprot.org
Stuck? Just ask – active help and support team
Feedback – if you find something incorrect, outdated, missing etc please tell us.

24. Introduction to Protein Signatures & InterPro
Introduction to InterPro

25. Protein Signatures Protein Signature =
an amino acid sequence (not necessarily consecutive) associated with a protein characteristic.
Basically introduce the concept of protein signatures
Introduction to InterPro

26. What value are signatures?
Better at finding proteins with common function
Find more distant homologues than BLAST
Better at finding proteins with common function
Classification of proteins
Associate proteins that share: Function
Domains
Sequence
Structure
Annotation of protein sequences
Define conserved regions of a protein
e.g. location and type of domains
key structural or functional sites

27. How are protein signatures made?
Protein family/domain
Build model
Search
Multiple sequence alignment
Significant matches
ITWKGPVCGLDGKTYRNECALL
AVPRSPVCGSDDVTYANECELK
SVPRSPVCGSDGVTYGTECDLK
HPPPGPVCGTDGLTYDNRCELR
E-value 1e-49
E-value 3e-42
E-value 5e-39
E-value 6e-10
Protein
signature
Refine
Introduction to InterPro

28. Types of Protein signatures
(sequence based)
Multiple protein alignment

29. Types of Protein signatures
(sequence based)
Single motif methods
Regular expression patterns
x = any AA
( ) = number of AAs
Must be this
C - C - {P} - x(2) - C - [STDNEKPI] - C
{ } = cannot be..
[ ] = any of

30. Types of Protein signatures
(sequence based)
Single motif methods
Regular expression patterns 1 2 3
Multiple motif methods
Identity matrices
Fingerprints

31. Types of Protein signatures Regular expression patterns
(sequence based)
Single motif methods
Regular expression patterns
Full domain alignment methods
Profiles
(Profile Library) M1 M2 M3 M4 I1 I2 I3 D2 D3
Multiple motif methods
Hidden Markov Models
Mathematical model of amino acid probability
Identity matrices
Fingerprints

32. CONTRIBUTING MEMBER DATA BASES
Models built on either sequence or structural alignments
Each MDB has its own focus
Hidden Markov Models
Finger-
Prints
Profiles
Patterns
Sequence
Clusters
InterPro uses signatures from several different databases (referred to as member databases) to predict information about proteins, such as possible function and the potential location of functionally important sites and domains.
Each member database creates signatures in different ways: some groups build them from manually-created sequence alignments, some use automatic processes with some human input and correction and others build their signatures entirely automatically. The signatures are represented using a variety of different model types (HMMs, Profiles, Regular Expressions, etc.)
The member databases all have their own particular niche or focus; at InterPro we aim to be a combination of their individual strengths. To do this we integrate signatures from the member databases that represent the same protein family, domain or site into a single InterPro entry. We check the biological accuracy of the individual signatures and add concise information about the signatures and the types of proteins they match, including consistent names, descriptive abstracts (with links to original publications) and GO terms.
Protein features (active sites…)
Prediction of conserved domains
Structural Domains
Functional annotation of families/domains

33. Database Basis Institution Focus URL Built from Pfam HMM
Sanger Institute
Sequence alignment
Family & Domain based on conserved sequence
Gene3D
UCL
Structure alignment
Structural Domain
Superfamily
Uni. of Bristol
Evolutionary domain relationships
SMART
EMBL Heidelberg
Functional domain annotation
TIGRFAM
J. Craig Venter Inst.
Microbial Functional Family Classification
Panther
Uni. S. California
Family functional classification
PIRSF
PIR, Georgetown, Washington D.C.
Functional classification
PRINTS
Fingerprints
Uni. of Manchester
PROSITE
Patterns & Profiles
SIB
Functional annotation
HAMAP
Profiles
Microbial protein family classification
ProDom
Sequence clustering
PRABI : Rhône-Alpes Bioinformatics Center
Conserved domain prediction

34. Foundations of InterPro
Integration of signatures
InterPro
Manual curation
Master headline

35. InterPro Entry Groups similar signature together
Links related signatures
Adds extensive annotation
Linked to other databases
Structural information and viewers
Master headline

36. Applies to domains and families
Link related signatures - relationships
Parent - Child
(subgroup of more closely related proteins) *
PFAM
(100)
Protein kinase
PFAM
(75)
(100)
SMART
Protein kinase
Serine kinase
PFAM
Protein kinase
SMART
PROSITE
Serine kinase
Tyrosine kinase
Parent
Children
Applies to domains and families
PROSITE
(25)
Tyrosine kinase
No proteins in common
SMART
PROSITE
Master headline

37. The InterPro entry types
Proteins share a common evolutionary origin, as reflected in their related functions, sequences or structure
Biological units with defined boundaries
Short sequences typically repeated within a protein
PTM
Active
Site
Binding
Conserved
Master headline

38. InterPro Search
protein ID
wwwdev.ebi.ac.uk/interpro

39. Unintegrated signatures
InterPro Search Results
Family
Link to PDBe
Domains and sites
Unintegrated signatures
Structural data

40. InterProScan – Searching New Sequence
Additional options
Paste in unknown sequence
wwwdev.ebi.ac.uk/interpro

41. Links to signature databases
InterProScan New Search Results
Link to InterPro entry
Links to signature databases
MENTION ALSO SUMMARY TABLE

42. Contact and help We need your feedback! missing/additional references
reporting problems
requests
help - questions answered
NOTE: Worth mentioning than sometimes we can not infer function, only homology to unknown proteins: is the case of Proteins of unknown function and Domains of unknown function. 42

43. Thanks for your attention