Anastasia Nikolskaya

1

COMPLEMENTING GENE ONTOLOGY WITH PIRSF
CLASSIFICATION-BASED PROTEIN ONTOLOGY

• Anastasia Nikolskaya
• PIR (Protein Information Resource)
• Georgetown University Medical Center
• www.uniprot.org http//pir.georgetown.edu/

2
Why Protein Classification?

• Automatic annotation of protein sequences based
  on protein families (propagation of annotation)
• Systematic correction of annotation errors
• Protein name standardization in UniProt
• Functional predictions for uncharacterized
  protein families

3
PIRSF Classification System

• PIRSF A network structure with hierarchies from
  Superfamilies to Subfamilies reflects
  evolutionary relationships of full-length
  proteins
• Definitions
• Basic unit Homeomorphic Family
• Homologous (Common Ancestry) Inferred by
  sequence similarity
• Homeomorphic Full-length sequence similarity and
  common domain architecture
• Network Structure Flexible number of levels with
  varying degrees of sequence conservation
• Advantages
• Annotation of both generic biochemical and
  specific biological functions
• Accurate propagation of annotation and
  development of standardized protein nomenclature
  and ontology

4
Levels of protein classification
5
PIRSF Classification System
A protein may be assigned to only one
homeomorphic family, which may have zero or more
child nodes and zero or more parent nodes. Each
homeomorphic family may have as many domain
superfamily parents as its members have domains.
6
PIRSF Classification System
A protein may be assigned to only one
homeomorphic family, which may have zero or more
child nodes and zero or more parent nodes. Each
homeomorphic family may have as many domain
superfamily parents as its members have domains.
SF500001 stimulates trophoblast
migration SF500002 stimulates proliferation of
prostate cancer cells SF500003
anti-proliferative and pro-apoptotic effects on
cancer cells SF500004 inhibitor of IGF
SF500005 stimulates bone formation SF500006
inhibitor of IGF-II
7
Creation and curation of PIRSFs
UniProt proteins
New proteins
Automatic Procedure
Unassigned proteins

• Computer-Generated (Uncurated) Clusters (36,000
  PIRSFs)
• Preliminary Curation (5,000 PIRSFs)
• Membership
• Signature Domains
• Full Curation (1,300 PIRSFs)
• Family Name with evidence tag
• Description, Bibliography

Automatic clustering
Preliminary Homeomorphic Families
Orphans
Map domains on Families
Automatic placement
Merge/split clusters
Add/remove members
Computer-assisted Manual Curation
Curated Homeomorphic Families
Name, refs, abstract, domain arch.
Final Homeomorphic Families
Protein name rule/site rule
Create hierarchies (superfamilies/subfamilies)
Build and test HMMs
8
PIRSF-Based Protein Annotation in UniProt
UniProt is developing protein name standards and
guidelines Classification of proteins into
families provides a convenient and accurate
mechanism to propagate curated information to
individual protein members

• Rule-Based annotation system using curated
  PIRSFs
• Site Rules (PIRSR) Position-Specific Site
  Features (active sites, binding sites, modified
  sites, other functional sites)
• Name Rules (PIRNR) transfer name from PIRSF to
  individual proteins (define a subgroup if
  necessary)
• Protein Name (may differ from family name),
  synonyms, acronyms
• EC
• Misnomers
• GO Terms (homeomorphic family-based, propagatable
  GO annotation)
• Function

9
PIRSF-Based Protein Ontology

• PIRSF family hierarchy is based on evolutionary
  relationships
• Standardized PIRSF family names
• Network structure (in DAG) for PIRSF family
  classification system

10
PIRSF to GO Mapping

• PIRSF to GO mapping provides a link between GO
  concepts and protein objects
• Mapped 5500 curated PIRSF homeomorphic families
  and subfamilies to the GO hierarchy

DynGO viewerHongfang Liu , University of Maryland

• Superimpose GO and PIRSF hierarchies
• Bidirectional display (GO-centric or
  PIRSF-centric views)

11
Protein Ontology Can Complement GO

• Expanding a Node
• Identification of GO subtrees that need expansion
  if GO concepts are too broad
• 67 of curated PIRSF families and subfamilies
  map to GO leaf nodes
• Among these, 2209 PIRSFs have shared GO leaf
  nodes (many PIRSFs to 1 GO leaf)
• Example PIRSF001969 vs PIRSF018239 and
  PIRSF036495 High- vs low-affinity IGF binding
• Identification of missing GO nodes

12
Protein Ontology Can Complement GO
Identification of Missing GO Nodes (higher
levels)
13
Protein Ontology Can Complement GO

• Linking Function, Biological Process, and
  Cellular Component through a Protein Object Based
  on Protein Annotations

• Mechanism to examine the relationships between
  the three GO ontologies based on the shared
  annotations at different protein family levels
• Example molecular function estrogen receptor
  activity and biological process signal
  transduction ,estrogen receptor signaling
  pathway

14
PIRSF Protein Classification a link between GO
and protein objects

• Annotation Quality
• Annotation of biological function of whole
  proteins
• Annotation of uncharacterized hypothetical
  proteins
• Correction of annotation errors and
  underannotations
• Standardization of Protein Names
• PIRSF to GO mapping provides a link between GO
  sub-ontologies and protein objects

15
PIRSF-based Protein Ontology Can Complement GO

• Identification of GO subtrees that need expansion
  if GO concepts are too broad
• Comprehensive classification of related protein
  families in PIRSF can help in identification of
  missing GO nodes when entire groups of PIRSF
  superfamilies or families cannot be mapped to
  existing GO terms
• Mechanism to examine the relationships between
  the three GO ontologies (molecular function,
  biological process, and cellular component), as
  well as between GO concepts, based on the shared
  annotations at different protein family levels

16
Acknowledgements

• Hongfang Liu , University of Maryland
• Judith Blake, The Jackson Laboratory
• Dr. Cathy Wu, Director
• Protein Classification team
• Dr. Winona Barker Dr. Lai-Su Yeh Dr.
  Anastasia Nikolskaya
• Dr. Darren Natale Dr. Zhangzhi Hu Dr. Raja
  Mazumder
• Dr. CR Vinayaka Dr. Xianying Wei Dr. Sona
  Vasudevan
• Informatics team
• Dr. Hongzhan Huang Baris Suzek, M.S.
  Sehee Chung, M.S.
• Dr. Leslie Arminski Dr. Hsing-Kuo Hua
  Yongxing Chen, M.S.
• Jing Zhang, M.S. Amar Kalelkar
• Students
• Christina Fang Vincent Hormoso Natalia
  Petrova Jorge Castro-Alvear

http//pir.georgetown.edu/
PIR Team
UniProt (SwissProt, TrEMBL, PIR)
www.uniprot.org