Sequence databases and retrieval systems

1

• Sequence databases and retrieval systems
• Guy Perrière
• Pôle Bioinformatique Lyonnais
• Laboratoire de Biométrie et Biologie Évolutive
• UMR CNRS n 5558
• Université Claude Bernard Lyon 1

2
In the beginning

• First paper compilation in 1965 (Atlas of Protein
  Sequences).
• Development of real databanks at the begin-ning
  of the 80s
• Fast access.
• Make possible analyses that require a lot of
  data
• Codon usage.
• Molecular phylogeny.

3
General databanks

• Nucleotide sequences
• EMBL/GenBank/DDBJ.
• Protein sequences
• Simple translations of coding regions
• GenPept.
• TrEMBL.
• Systems containing specific data
• SWISS-PROT.
• PIR.

4
EMBL

• Created in 1980 at the European Molecular Biology
  Laboratory in Heidelberg.
• Maintained since 1994 at the European
  Bio-informatic Institute (EBI) in Cambridge.
• Web server
• http//www.ebi.ac.uk/embl

5
GenBank

• Set up in 1979 at the Los Alamos National
  Laboratory (LANL) in Los Alamos.
• Maintained since 1992 at the National Cen-ter for
  Biotechnology Information (NCBI) in Bethesda.
• Web server
• http//www.ncbi.nlm.nih.gov/Genbank/index.html

6
DDBJ

• Started its activities in 1984 at the National
  Institute of Genetics (NIG) in Mishima.
• Since then, still maintained in this institute by
  the team of Takashi Gojobori.
• Web server
• http//www.ddbj.nig.ac.jp

7
Nucleotide sequences

• Data mainly provided by direct submissions from
  the authors.
• Submissions are made through the Internet
• Web forms.
• Email.
• The sequences are exchanged between the three
  centers on a daily basis
• The content of the banks is identical.

8
Data growth
EMBL GenBank NBRF/PIR SWISS-PROT
log(Nb. of residues)
9
GenBank size (May 2001)

• 12.7?109 nucleotides.
• 11.8?106 sequences.
• 690,323 genes (proteins and RNA).
• 242,616 bibliographic references.
• 47.8 gigabytes on disc.
• Growth of 280 in 12 months.
• 24-36 h to download the whole GenBank files from
  NCBI.

10
Taxonomic sampling

• There are 72,000 species for which at least one
  sequence is available.
• Nine species (0.01) correspond to 85 of the
  total.
• 18,000 species are represented by only one
  sequence!

The nine species the most represented in GenBank
11
Distribution format

• The banks are distributed as a set of text files
  ( 215 for GenBank).
• A file contains sequences corresponding to
• A given taxon (e.g., bacteria, invertebrates,
  mammals).
• A given class of sequences (EST, HTG, GSS).
• Inside a file, each sequence is called an entry.

12
Entries structure

• Informations are introduced into structured
  fields.
• The format is different between EMBL and
  GenBank/DDBJ.
• The data introduced in the three databanks are
  identical.

13
ID, AC, SV and DT fields
Contain identifyers as well as the creation and
the last modification dates for the entries. ID
BSAMYL standard DNA PRO 2680 BP. XX AC
V00101 J01547 XX SV V00101.1 XX DT
13-JUL-1983 (Rel. 03, Created) DT 12-NOV-1996
(Rel. 49, Last updated, Version 11)
14
DE, KW, OS and OC fields
Contain general information on sequences
(defini-tion, keywords, taxonomy). DE Bacillus
subtilis amylase gene. XX KW amyE gene
amylase amylase-alpha KW regulatory region
signal peptide. XX OS Bacillus subtilis OC
Bacteria Firmicutes Bacillus/Clostridium
group OS Bacillus/Staphylococcus group
Bacillus.
15
RN, RX, RA and RT fields
Contain informations related to bibliographic
refe-rences. RN 1 RP 1-2680 RX MEDLINE
83143299. RA Yang M., Galizzi, A., Henner,
D.J. RT "Nucleotide sequence of the amylase
gene from RT Bacillus subtilis" RL Nucleic
Acids Res. 11237-249(1983).
16
FT field
Contains the descriptions of functional regions
described by the qualifyers. FT promoter
369..374 FT /note"put. promoter
sequence P2 3 (amyR1)" FT RBS
414..419 FT /note"rRNA-binding site
rbs-1 3" FT CDS 498..2480 FT
/gene"amyE" FT /db_xref"SWISS-PROT
P00691" FT /product"alpha-amylase
precursor" FT /EC_number"3.2.1.1  FT
/protein_id"CAA23437.1" FT
/translation"MFAKRFKTSLLPLFAGFLLLFHLVLAGPAA FT
ASAETANKSNELTAPSIKSGTILHAWNWSFNTLKHNM
KDIHDAG ...
17
Intron/exon structure
FT CDS join(242..610,3397..3542,5100..535
1) FT /codon_start1 FT
/db_xref"SWISS-PROTP01308" FT
/note"precursor" FT /gene"INS" FT
/product"insulin" ...
18
SQ field
Contains the sequence iself SQ Sequence 2680
BP 825 A 520 C 642 G 693 T 0 other
gctcatgccg agaatagaca ccaaagaaga actgtaaaaa
cgggtgaagc agcagcgaat 60 agaatcaatt
gcttgcgcct ttgcggtagt ggtgcttacg atgtacgaca
gggggattcc 120 ccatacattc ttcgcttggc
tgaaaatgat tcttcttttt atcgtctgcg gcggcgttct
180 gtttctgctt cggtatgtga ttgtgaagct
ggcttacaga agagcggtaa aagaagaaat 240
(...) gatggtttct tttttgttca taaatcagac
aaaacttttc tcttgcaaaa gtttgtgaag 2580
tgttgcacaa tataaatgtg aaatacttca caaacaaaaa
gacatcaaag agaaacatac 2640 cctgcaagga
tgctgatatt gtctgcattt gcgccggagc
2680 //
19
Errors in databanks

• There are a lot of errors in the nucleotide
  sequence databanks
• For the annotations, the free submission of
  entries involves
• Inaccuracies, omissions, and even mistakes.
• Inconsistencies between some fields.
• In the sequences themselves
• Sequencing erors.
• Compression, gel reading.
• Cloning vectors inserted.

20
Redundancy

• Another major pro-blem is redundancy.
• A lot of entries are partially or entirely
  duplicated
• 20 of vertebrate se-quences in GenBank.
• Duplicated entries are often different in their
  sequence.

21
Variations in duplicates

• It is often impossible to decide whether a
  difference between two duplicates is due to
• Polymorphism.
• Sequencing error.
• True gene duplication.
• And what to do when annotations are diffe-rent or
  even contradictory?

22
Protein sequences

• Translation of Coding DNA Sequences (CDS) from
  EMBL/GenBank/DDBJ.
• Consultation of publications or patents.
• Small number of direct protein submissions by
  authors.
• Integration of specific annotations only for the
  true protein sequences databanks.

23
SWISS-PROT

• Created by Amos Bairoch in 1986 at the
  Departement of Medical Biochemistry in Geneva.
• Maintained by the Swiss Institute of
  Bio-informatics (SIB) and funded by GeneBio.
• Web server
• http//www.expasy.ch/sprot/sprot-top.html

24
SWISS-PROT characteristics

• Almost no redundancies.
• Cross-references with 55 other databanks.
• High-quality annotations
• Systematic control by a team of annotators.
• Help of a set of 204 benevolent experts.
• Embedded in a complete environment devoted to
  proteins.

25
Annotations

• Protein function.
• Post-translational modifications.
• Structural or functional domains.
• Secondary and quaternary structures.
• Similarities with other proteins.
• Conflicts between positions for CDS.

26
Associated databanks

• TrEMBL, built using only annotated CDS from EMBL.
• ENZYME, for the international enzyme
  nomenclature.
• PROSITE, for biologically significant sites,
  patterns and profiles.
• SWISS-2DPAGE, for two-dimensional polyacrylamide
  gel electrophoresis maps.

27
PIR

• PIR (The Protein Information Resource) was
  created by M.O. Dayhoff in 1965.
• Aims
• To provide exhaustive and non-redundant protein
  data.
• To give a classification using taxonomic and
  similarity data
• Grouping of entries into super-families, families
  and sub-families.

28
Data maintenance

• Three organisms collect and organize the data
  introduced in PIR
• The National Biomedical Research Foundation
  (NBRF) in the United States.
• The Martinsried Institute for Protein Sequence
  (MIPS) in Germany.
• The Japan International Protein Sequence
  Information Database (JIPID) in Japan.

29
Results

• The exhaustivity is not better than what is
  obtained with SWISS-PROTTrEMBL.
• Still contains a lot of redundancies.
• Lower quality for the annotations.
• Low number of cross-references.

30
Specialized databanks

• A lot of specialized databanks has been
  developed, which are devoted to
• Complete genomes.
• Families of homologous genes.
• Non-sequence data.
• These systems are under the responsibility of
  curators
• Data quality and homogeneity control.

31
Complete genomes

• There is a large number of databanks devo-ted to
  peculiar organisms.
• These banks are associated to sequencing or
  mapping projects.
• For some model organisms there are often
  concurrent systems.

32
Examples
33
Gene families databanks

• Built thanks to automated procedures
• Similarity search between sets of proteins
  (BLASTP, FASTP, Smith-Waterman).
• Clustering into homologous families using
  similarity criteria.
• Include various data
• Protein (and sometimes nucleotide) sequences.
• Multiple alignments and trees.
• Taxonomy.

34
ProtFam

• Developed at MIPS.
• Built with PIR sequences.
• Includes four levels of classification
• Superfamilies (based on function and similarity
  criteria).
• Families (50 similarity).
• Subfamilies (80 similarity).
• Entries (95 similarity).

35
ProtFAm characteristics

• Allows to visualize alignments and dendro-grams
  for the families.
• Integrates Pfam domains.
• Allows users to classify their own protein
  sequences.
• Web server
• http//www.mips.biochem.mpg.de

36
ProtoMap

• Developed at the Department of Biological
  Chemistry from The Hebrew University of
  Jerusalem.
• Built with SWISS-PROT sequences.
• Uses three similarity measures for sequences
  (BLASTP, FASTP and Smith-Waterman).

37
ProtoMap characteristics

• Alignments and trees are visualized thanks to
  Java applets.
• Possibility for the users to submit their own
  sequences in a way to classify them.
• No domain data but integrates the possibility to
  visualize low-similarity relationships.
• Web server
• http//www.protomap.cs.huji.ac.il

38
Specialized systems

• HOVERGEN (Homologous Vertebrate Genes Database)
  for vertebrates
• Based on GenBank CDS.
• HOBACGEN (Homologous Bacterial Genes Database)
  for prokaryotes and yeast
• Based on SWISS-PROT/TrEMBL.
• HOBACGEN-CG for completely sequenced genomes
• Based on SWISS-PROT/TrEMBL.

39
Other specialized systems

• COG (Clusters of Orthologous Groups), also for
  complete genomes
• Based on GenBank CDS.
• NuReBase (Nuclear Receptors Database) for
  mammalian nuclear receptors
• Based on EMBL CDS.
• RTKdb (Tyrosine Kinase Receptors)
• Based on EMBL CDS.

40
Are COGs real orthologs?
Escherichia coli Bacillus subtilis Pseudomonas Aer
uginosa Vibrio cholerae Synechocystis sp.
Glutamate synthase large subunit
41
HOBACGEN

• Integrates protein and nucleic sequences as well
  as multiple alignments and trees.
• Is based upon a client/server architecture.
• Client software is distributed as well as the
  server structure (including all sequences).
• Web server
• http//pbil.univ-lyon1.fr/databases/hobacgen.html

42
Similarities search
?
SWISS-PROT/TrEMBL sequences
43
Segments selection
44
Families assembling
45
Alignments and trees
Rooting by mid-point
46
Domain structure
6PG1_YEAST
6PGD_CANAL
6PGD_SOYBN
6PG2_BACSU
O32911_MYCLR
P95165_MYCTU
6PGD_CERCA
Q40311_MEDSA
Y770_MYCTU
Y229_SYNY3
ProDom domains for the 6PGD family
47
Examples

• Domains (Blocks, Domo, Pfam, ProDom, SBASE).
• Sites, patterns and profiles (PRINTS, PRO-SITE).
• The InterPro databank gather all available data
  on domains and patterns with a known biological
  function.

48
Non-sequence data
49
Data retrieval

• Made mainly through Internet access
• With client software (e.g., Entrez, HobacFetch).
• By remote connections to servers providing
  on-line access to the banks (INFOBIOGEN).
• Using World-Wide Web servers and browsers
  (Netscape, Internet Explorer, Lynx, etc.)

50
Advantages and limitations

• Users do not have to cope with the usual
  databases problems
• Storing of large amounts of data.
• Daily updates.
• Software upgrades.
• Simplicity of use.
• Net access is sometimes very slow at peak hours
• Use your local servers instead of NCBI!!!

51
Retrieval systems

• Direct access to functional regions described in
  the features (CDS, tRNA, rRNA).
• Selection of entries using various criteria
• Sequence names and accession numbers.
• Bibliographic references.
• Keywords.
• Taxonomy.
• Publication date.
• Organelle, host.

52
Query

• Developed at the Laboratoire de Biométrie et
  Biologie Évolutive by Gouy et al. (1985).
• Graphical interface distributed along with the
  databases themselves.
• Web access at Pôle Bioinformatique Lyon-nais
  (PBIL)
• http//pbil.univ-lyon1.fr/search/query.html

53
Characteristics

• Allows to query any bank in PIR, SWISS-PROT,
  EMBL, or GenBank/DDBJ formats.
• Keywords and species browsing.
• Complex queries.
• Links with sequence analysis programs on the Web
  server (alignment, codon usage).

54
SRS

• Public version developed at EMBL by Etzold and
  Argos (1993).
• Presently available on the different Web servers
  belonging to EMBnet
• INFOBIOGEN (France).
• EBI (England).
• DKFZ (Germany).

55
Characteristics

• Database index built thanks to the use of ODD
  (Object Design and Definition).
• More than 250 databanks have been indexed and are
  accessible through 35 SRS servers.
• Allows queries on different banks simul-taneously.

56
Databanks interconnection
57
Entrez

• Developed by Schuler et al. (1996) at NCBI.
• Allows to query only databases that are made in
  the USA
• GenBank, GenPept, NR, MMDB, MEDLINE.
• Access through client software (Unix, Mac or
  Windows) or Web server
• http//www.ncbi.nlm.nih.gov

58
Characteristics

• Introduce the concept of neighbours between
  sequences, references and structures.
• Sequence neighbours are establish using
  similarity criteria.
• No access to multiple alignments.