Intelligent Curation Using Ontologies

1
Intelligent Curation Using Ontologies

• K.Wolstencroft

2
Introduction

• Developing an automated system for extracting and
  classifying proteins from newly sequenced genomes
  
• Background
• Architecture
• Advantages

3
Motivation

• Genome sequencing techniques greatly improved
• More whole genomes are being sequenced quickly -
  lots of data being generated
• Without analysis and classification sequences
  are simply a series of letters
• Therefore, data analysis is now the rate-limiting
  step

4
Why Classify?

• Classification and curation of a genome is the
  first step in understanding the processes and
  functions happening in an organism
• Classification enables comparative genomic
  studies - what is already known in other
  organisms
• The similarities and differences between
  processes and functions in related organisms
  often provide the greatest insight into the
  biology

5
BackgroundDNA to Proteins

• Genome sequencing produces DNA sequences
• DNA blueprint of an organism
• DNA encodes complex molecules mostly proteins
• Proteins are the functional molecules of a cell

6
Proteins

• Complex molecules constructed from sequences of
  amino acids
• 20 different amino acids with different chemical
  properties

7
Proteins Primary Structure

• Amino acid sequences can be represented as a
  series of single letters
• gt1A5Y_ PROTEIN TYROSINE PHOSPHATASE 1B
• MEMEKEFEQIDKSGSWAAIYQDIRHEASDFPCRVAKLPKNKNRNRYRDV
  SPFDHSRIKLHQEDNDYINASLIKMEEAQRSYILTQGPLPNTCGHFWEMV
  WEQKSRGVVMLNRVMEKGSLKCAQYWPQKEEKEMIFEDTNLKLTLISEDI
  KSYYTVRQLELENLTTQETREILHFHYTTWPDFGVPESPASFLNFLFKVR
  ESGSLSPEHGPVVVHXSAGIGRSGTFCLADTCLLLMDKRKDPSSVDIKKV
  LLDMRKFRMGLIATAEQLRFSYLAVIEGAKFIMGDSSVQDQWKELSHEDL
  EPPPEHIPPPPRPPKRILEPHNGKCREFFPN

8
ProteinsTertiary Structure

• Sequence determines structure

9
Searching for Features

• The relationship between amino acid sequence and
  eventual protein structure means that we can
  search for distinct structural (and functional)
  domains within the sequence
• Domains could be several amino acids long or
  could span most of the protein

10
Example

• A search of the linear sequence of protein
  tyrosine phosphatase type K identified 9
  functional domains
• gtuniprotQ15262PTPK_HUMAN Receptor-type
  protein-tyrosine phosphatase kappa precursor (EC
  3.1.3.48) (R-PTP-kappa).
• MDTTAAAALPAFVALLLLSPWPLLGSAQGQFSAGGCTFDDGPGACDYHQD
  LYDDFEWVHV
• SAQEPHYLPPEMPQGSYMIVDSSDHDPGEKARLQLPTMKENDTHCIDFSY
  LLYSQKGLNP
• GTLNILVRVNKGPLANPIWNVTGFTGRDWLRAELAVSSFWPNEYQVIFEA
  EVSGGRSGYI
• AIDDIQVLSYPCDKSPHFLRLGDVEVNAGQNATFQCIATGRDAVHNKLWL
  QRRNGEDIPV..

11
Human Expert Annotation

• Bioinformaticians use a series of tools to
  identify functional domains
• Similarity searching, domain/motif identification
• Tools include BLAST / INTERPRO
• Tools simply show presence of domains
• Use expert knowledge to classify proteins
  according to domain arrangements
• Presence / order / number of each important
• Can an ontology be used to capture this knowledge
  to the standard of a human annotator?

12
Protein Family Classification

• Proteins divided into broad functional classes
  Protein Families
• Often diagnostic domains/motif signify family
  membership
• Initial Study focuses on the protein phosphatase
  family

13
The Protein Phosphatases

• large superfamily of proteins involved in the
  removal of phosphate groups from molecules
• Important proteins in almost all cellular
  processes
• Involved in diseases diabetes and cancer
• human phosphatases well characterised

14
Characterisation allows classification

• Diagnostic phosphatase domains/motifs
  sufficient for membership of the protein
  phosphatase superfamily
• Other motifs determine a proteins place within
  the family
• This human expert knowledge can be captured and
  incorporated into the model if the domain
  organisations are represented in a formal DL OWL
  ontology

15
Protein Functional Domains
Andersen et al (2001) Mol. Cell. Biol. 21 7117-36
16
Determining Class Definitions

• R2A
• Contains 2 protein tyrosine phosphatase domains
• Contains 1 transmembrane domain
• Contains 4 fibronectin domains
• Contains 1 immunoglobulin domain
• Contains 1 MAM domain
• Contains 1 cadherin-like domain

17
Protégé OWL Modelling
18
Requirements

• Extract phosphatase sequences from rest of
  protein sequences from a whole genome
• Identify the domains present in each
• Compare these sequences to the formal ontology
  descriptions
• Classify each protein instance to a place in the
  hierarchy

19
Technology
OWL DL ontology
Instance Store
myGrid Workflow
Classified Protein Phosphatases
Raw protein sequences
Reasoner (racer)
20
myGrid Workflow

• extract sequences from whole genome
• perform simple filtering patmatdb
• performs InterproScan to determine domain
  architecture
• transform the InterproScan results into abstract
  OWL instance descriptions

21
myGrid Workflow
22
InterproScan Results
23
Conversion to abstract OWL format

• restriction(lthttp//www.owl-ontologies.com/unnamed
  .owlcontainsDomainIPR000340gt cardinality(1))
• restriction(lthttp//www.owl-ontologies.com/unnamed
  .owlcontainsDomainIPR001763gt cardinality(1))
• restriction(lthttp//www.owl-ontologies.com/unnamed
  .owlcontainsDomainIPR000387gt cardinality(1))

24
Instance Store

• Instance Store enables reasoning over individuals
• Can support much higher numbers of individuals
• OWL ontology is loaded into the instance store
• A DL reasoner (racer) is used to compare
  individuals to the OWL ontology definitions

25
Instance Store
26
Example Instances

• Protein Individual
• Dual Specificity Phosphatase DUSE
• restriction(lthttp//www.owl-ontologies.com/unnamed
  .owlcontainsDomainIPR000340gt cardinality(1))
• restriction(lthttp//www.owl-ontologies.com/unnamed
  .owlcontainsDomainIPR000387gt cardinality(1))

• Ontology Definition of Dual Specificity
  Phosphatase
• containsDomain IPR000340
• Necessary and Sufficient for class membership
• Also inherits
• containsDomain IPR000387 from Parent Class PTP

27
So Far..

• Human phosphatases have been classified using the
  system
• The ontology classification performed equally
  well as expert classification
• The ontology system refined classification
• - DUSC contains zinc finger domain
• Characterised and conserved but not in
  classification
• - DUSA contains a disintegrin domain
• previously uncharacterised evolutionarily
  conserved

28
Aspergillus fumigatus

• Phosphatase compliment very different from human
• gt100 human lt50 A.fumigatus
• Whole subfamilies missing
• Different fungi-specific phosphorylation
  pathways?
• No requirement for tissue-specific variations?
• Novel serine/threonine phosphatase with homeobox
• conserved in aspergillus and closely related
  species, but not in any other

29
Conclusions

• Using ontology allows automated classification to
  reach the standard of human expert annotation
• Reasoning capabilities allow interpretation of
  domain organisation
• Produces interesting biological questions
• Allows fast, efficient comparative genomics
  studies
• System currently describes protein phosphatases -
  but possible to expand to other protein families

30
Acknowledgements

• Group myGrid
• PhD Supervisors Andy Brass, Robert Stevens
• Phosphatase Biologist Lydia Tabernero
• Ontogrid and NIBHI