1. Ontologies, Databases, Knowledgebases: How should they interoperate?
Judith Blake, Ph.D.
The Jackson Laboratory
Database: organized body of related information; a structured collection of records or data
Knowledgebase: facts + inferences

2. Thesis The Mouse Genome Informatics (MGI) system
provides a model for interoperablity that
incorporates the use of ontologies,
depends upon the interconnection among databases, and
Supports integration of data from multiple data sources
This may provide model for PRO objectives to support connections between PRO and disease representations

3. Mouse Genome Informatics (MGI)
MGI’s primary mission is to facilitate the use of mouse as a model for human biology by providing integrated access to data on the genetics, genomics, and biology of the laboratory mouse.
variants &
polymorphisms
expression
strain geneaology
Hermansky-Pudlak syndrome
Mouse model & human phenotype
sequence
genome location
tumors
gene function
mouse/human
orthologs & maps
Information content spans from sequence to phenotype/disease

4. Automated (mostly) Data Integration (Loads)
Clones
EG mouse
RPCI GO
UniProt
MGC MP
Associations
Vocabularies
DFCI
Anatomy
DoTS
Interpro
NIA
OMIM
Unigene
PIRSF
TreeFam
Add: Vega?
Homologene
Gene trap
Entrezgene chimp & dog
Remove: Riken
Change TIGR to DFSR (or whatever it is)
Annotation
Gene traps
MGI
GenBank
EG chimp
EG dog
RefSeq
Sequences
EG rat
UniProt
EG human
DFCIseq
HCOP
DoTSseq
Homologene
NIAseq
Non-mouse
dbSNP
NCBI
VEGA
SNP db
UniSTS
Gene models and coordinates
Ensembl
microRNAs

5. Mouse Genome Informatics Controlled vocabularies and ontologies
GO - Gene Ontology (GO)
PRO - Protein Ontology (PRO)
MP - Mouse Phenotype Ontology
MA - Mouse Anatomy (GXD)
CL- Cell Type Ontology
Mouse gene and strain nomenclature
SO - Sequence Ontology
RO - Relations Ontology
ECO - Evidence Code ontology
Integration: Controlled Vocabularies and Ontologies

6. MGI Operating Principles
Data integration is key to comprehensive access to mouse genome, functional, mouse model, and comparative data allows the data to be evaluated in new contexts
Supports robust access to comprehensive information
Permits efficient access to related resources
Standards are key to data integration
Nomenclature
Standardized gene nomenclature, keywords, etc.
Knowledge representation
Gene Ontology (GO)
Mammalian Phenotype Ontology
Integration of Multi-Source Data
Depends on consistent entity tagging
Requires improvement of data storage structures
Necessitates ontology updates for data categories and context

7. Mouse Phenotypes and Disease Models Connects mouse and human phenotypes in studies of human disease processes
Mouse Crebbptm1Sis/Crebbp+
mutants showing skeletal formation defects.
Human Rubinstein-Taybi Syndrome 1 (OMIM:180849), caused by CREBBP mutation.
° mental retardation
° postnatal growth deficiency
° microcephaly
° broad thumbs & halluces
° dysmorphic facial features (beaked nose, high arched palate, characteristic grimacing)
° increased tumor risk 1

8. Diseases and Phenotypes
Diseases are described by signs and symptoms
Signs – things you can measure
Symptoms – things the patient notices
Signs are phenotypes
Diseases are characterized by phenotypes including the order, severity and duration with which they occur. A full model of disease takes into account dimensions of anatomy, time, severity, therapeutic responsiveness, outcomes etc. There is also a probabilistic element to an instance of the disease and a probabilistic association between phenotypic elements in one instance.
Diseases are not phenotypes ( although predisposition may be considered as such) but single phenotype diseases may be viewed as phenotypes, eg. osteoarthritis.
Paul Schofield, 2013

9. Status of Phenotype & Disease Data
May 2012
May 2013
May 2014
change this yr.
Phenotype terms in MP ontology
8,775
9,034
10,190
+1,156
Mutant alleles cataloged : total
: in mice
number of genes represented
targeted alleles
number of genes targeted
743, ,299
20,937
46,822
15,488
748,960
33,659
21,442
51,119
16,221
754,256
39,241
21,786
55,640
16,358
+5,296
+5,582
+344
+14,
Alleles w/ phenotype (MP) annotation
Genotypes with MP annotation
Total MP annotations
29,064
43,579
223,125
32,095
47,790
249,460
34,625
51,720
268,577
+2,530
+3,930
+19,117
Mouse genotypes modeling human disease
Human Diseases w/1 mouse model(s)
3,687
1,153
4,084
1,239
4,365
1,310
+281
+71
QTLs
4,696
4,715
4,835
+120

10. Objective
…make phenotype and disease model data robust and accessible to researchers and computational biologists
semantic consistency to enable complete data retrieval
integrated access to all phenotypic variation sources
(single-gene and genomic mutations, engineered mutations, QTLs, strains)
data on human disease correlation
access to mouse models from various approaches
- Genetic
- Phenotypic
- Genomic localization
- Computational

11. Annotating Disease to Genotype
Different alleles of a gene on the same background may/may not be disease models
The same alleles of a gene on different genetic backgrounds may/may not be disease model
Disease models are attached to genotype “objects”
Disease annotation consists of OMIM term, the data reference /source, and association type
OMIM term
129S1/Sv
Crouzon Syndrome
genotype
Fgfr2tm1Schl / Fgfr2+
phenotypic similarity to human
disease associated with ortholog
association type
Eswarakumar VP et al.,
PNAS USA 2006;103:
source 8

12. MGI
4,084 MGI Mouse Models
1,239 OMIM diseases (associated with)

13. Note chicken and zebrafish
Each associated human disease links to a Human Disease and Mouse Model Detail Page
Survival of motor neuron 1
Note chicken and zebrafish

14. Mouse Genome Informatics: Integrate Sequence with Biology
Nucleotide
Sequences
Nomenclature
Genome location
Strains
Polymorphisms
Orthology
Expression
Alleles
Mutant phenotypes
Function of gene products
Literature
Genome
Features
Genome
variation
Biological knowledge and
attributes in MGI
Protein
Sequences
Gene
predictions

15. Disease
Cell
Anatomy
Adapted from Schriml and Kibbe: ICBO submission 2013

16. Now with annotation extensions
positive regulation of transcription from pol II promoter in response to oxidative stress[GO: ]
protein localization to nucleus[GO: ]
cellular response to oxidative stress
[GO: ]
happens
during
sty1
pap1
has
input
has regulation
target
<anonymous
description>
<anonymous
description>
Key point: logically equivalent to an annotation to a term in the <anon desc> box, with the same links out. DB
Object
Term Ev
Ref
Extension
PomBase
sty1
SPAC24B11.06c
GO:
protein localization to nucleus
IMP
PMID: ..
happens_during(GO: ),
has_input(SPAC c)
pap1
SPAC c
GO:
has_regulation_target(…)

17. Annotation Extensions

18. MGI Modular Annotation Example–
Xirp1 is involved in the organization of the sarcomere
in a cardiac muscle cell (CL: ) of the myocardilum (MA: )
xin actin-binding repeat containing 1
Total number of MGI modular annotation units to proteins: 22,866
This does not include annotations to permanent cell lines

19. Summary of MGI Modular Annotations
part_of
9013
occurs_in
6298
regulates_o_occurs_in
2884
regulates_o_acts_on_population_of
1017
regulates_o_results_in_acquisition_of_features_of
967
results_in_acquisition_of_features_of
723
regulates_o_has_agent
537
regulates_o_has_participant
275
acts_on_population_of
259
results_in_movement_of
232
results_in_development_of
173
regulates_o_results_in_movement_of
156
results_in_specification_of 96
results_in_maturation_of 61
results_in_morphogenesis_of 48
has_agent 34
results_in_commitment_to 32
results_in_division_of 21
regulates_o_results_in_commitment_to 14
results_in_determination_of 13
regulates_o_results_in_specification_of 7
has_output_o_axis_of 5
regulates_o_results_in_development_of 1

20. Interaction Data in MGI …from catalog to context
Relationships among markers project
Explicit representation of relationships among genome features
Interaction explorer
Project initially focused on microRNAs
microRNA cluster membership
Predicted and validated microRNA targets
Curation of interaction data from the literature (Gene Ontology) and from specialized external informatics resources

21. Mouse_CCO is an application ontology built on experimental evidence-based annotations. The data drives the structure allowing a user to ‘discover’ connections. This diagram illustrates the generic template for the ontology.
Protein_mouse
PRO, UniProtKB
Gene_mouse
MGI
encodes
Allele_mouse
Genotype_mouse
has_variant
part_of
Gene_human
NCBI
orthologous_to
Phenotype_mouse
Disease_human
OMIM, DO
CCO_human
BioPortal
associated_with
Pathway_mouse
MouseCyc
described_by
Process
GO (BP)
Component
GO (CC)
Function
GO (MF)
participates_in
located_in
Anatomy_mouse
GXD, EMAP
expressed_in
effects
Cell_type CL
Mary Dolan

22. Mouse_CCO is populated using 1017 mouse genes annotated to GO ‘cell cycle’ along with all their annotations from MGI and several additional data resources. Here we show how the generic template is populated for Brca1.
orthologous_to
has_variant
Gene_human
BRCA1
Mouse gene: Brca1
(breast cancer 1)
Allele:
Brca1tm1Thl
part_of
described_by
encodes
Genotype:
Brca1tm1Thl/Brca1tm1Thl Waptm1(cre)Arge/0
129S1/Sv * C57BL/6J
id: CCO:B
name: BRCA1_HUMAN
Protein_mouse
VEGA model
OTTMUSP
participates_in
associated_with
Process
DNA repair
associated_with
mammary adenocarcinoma
Function
damaged DNA binding
participates_in
Component
BRCA1-BARD1 complex
located_in
expressed_in
effects
OMIM: Breast Cancer
TS28:
mammary gland
Mary Dolan
associated_with

23. Keys to Interoperability self-help mantras
Start where you are: from silos to networks
Identify shared interests: educated self promotion
Develop shared processes/applications
Discuss the ideal, implement the practical

24. Acknowledgements Gene Ontology Funding: NIH_NHGRI
Mike Cherry
Suzi Lewis
Paul Sternberg
Paul Thomas
Funding: NIH_NHGRI
Mouse Genome Informatics
Carol Bult
Janan Eppig
Jim Kadin
Joel Richardson
Martin Ringwald
MGI-GO-PRO team
Karen Christie
Mary Dolan
Harold Drabkin
David Hill
Li Ni
Dmitry Sitnikov