Ontologybased Links and Genomic Information Retrieval

1
Ontology-based Links and Genomic Information
Retrieval
Fabrice Camous September 2005 PhD Transfer
Talk Supervisor Stephen Blott and Alan
Smeaton School of Computing Dublin City
University Acknowledgements My research work is
funded by Enterprise Ireland under the Basic
Research Grants Scheme, project number
SC-2003-0047-Y.
2
Genomic Information
3
Talk Plan

• Links in Genomic Information.
• Background
• - Ontologies.
• - Network-based Similarity Measures.
• Our Work.
• Related Work.
• PhD Hypothesis.
• Future Experiments.

4
Genomic Information
Poorly Studied Gene
ctcggagggc accgcccacc gggg
?
Links?
Well Studied Gene
cagccgaggc ctcggagggc accg
Poorly Studied Gene
ctcggagggc accgcccacc gaga
?
Links?
Poorly Studied Gene
ctcggagaac accgcccacc gggg
Links?
?
5
Genomic Information
Explicit Links
(absolute links)
Citations
ctcggagggc accgcccacc gggg
Homologies (common ancestral history)
cagccgaggc ctcggagggc accg
cgggcgaaac ttcgccggct acca
Biomedical Literature (NCBI MEDLINE)
Cross-domain References
Nucleotide Sequence Database (NCBI GenBank)
6
Genomic Information
Implicit Links
(fuzzy links)
Our hypothesis ontology-based descriptions
provide additional information.
Ontology-based Similarity
Ontology-based Similarity
ctcggagggc accgcccacc gggg
Text-based Similarity
cagccgaggc ctcggagggc accg
Sequence Similarity
cgggcgaaac ttcgccggct acca
Biomedical Literature (NCBI MEDLINE)
Nucleotide Sequence Database (NCBI GenBank)
7
Genomic Information
Ontology-based Similarities
Ontology-based Similarity
Ontology-based Similarity
ctcggagggc accgcccacc gggg
cagccgaggc ctcggagggc accg
cgggcgaaac ttcgccggct acca
Biomedical Literature (NCBI MEDLINE)
Nucleotide Sequence Database (NCBI GenBank)
8
Genomic Information
Medical Subject Headings
Gene Ontology
-22,997 terms -Network set of 15
DAGs -Relationships broader/narrower than
-About 4,500 terms (April 2000) -Network set of
3 DAGs -Relationships is-a and part-of
Genetic Processes
Cell
Binding
Evolution
Gene Expression
Cell Division
Membrane
Nucleic Acid Binding
9
MeSH-based Representations
MeSH-based Implicit Link
PMID- 10506108 TI - Reduction of UV-induced skin
tumors in hairless mice by selective COX-2
inhibition. MH - Animals MH - Carcinoma,
Squamous Cell/enzymology/pathology/prevention
control MH - Cell Division MH - Cyclooxygenase
Inhibitors/therapeutic use MH - Female MH -
Immunohistochemistry MH - Isoenzymes/drug
effects/metabolism MH - Mice MH - Mice, Inbred
HRS MH - Neoplasms, Radiation-Induced/enzymology/
pathology/prevention control MH -
Prostaglandin-Endoperoxide Synthase/drug
effects/metabolism MH - Skin Neoplasms/enzymology
/pathology/prevention control MH -
Sulfonamides/therapeutic use MH - Support, U.S.
Gov't, P.H.S. MH - Ultraviolet Rays

• PMID- 10434051
• TI - Quantitative alterations of hyaluronan and
  dermatan sulfate in the
• hairless mouse dorsal skin exposed to
  chronic UV irradiation.
• MH - Animals
• MH - Chondroitin ABC Lyase
• MH - Collagenases
• MH - Deoxyribonucleases, Type II Site-Specific
• MH - Dermatan Sulfate/radiation effects
• MH - Disaccharides/analysis
• MH - Female
• MH - Histological Techniques
• MH - Hyaluronic Acid/radiation effects
• MH - Mice
• MH - Mice, Inbred HRS
• MH - Skin/chemistry/pathology/ radiation
  effects
• MH - Swine
• MH - Ultraviolet Rays

10
MeSH-based Representations
MeSH-based Implicit Link
PMID- 10506108 TI - Reduction of UV-induced skin
tumors in hairless mice by selective COX-2
inhibition. MH - Animals MH - Carcinoma,
Squamous Cell/enzymology/pathology/prevention
control MH - Cell Division MH - Cyclooxygenase
Inhibitors/therapeutic use MH - Female MH -
Immunohistochemistry MH - Isoenzymes/drug
effects/metabolism MH - Mice MH - Mice, Inbred
HRS MH - Neoplasms, Radiation-Induced/enzymology/
pathology/prevention control MH -
Prostaglandin-Endoperoxide Synthase/drug
effects/metabolism MH - Skin Neoplasms/enzymology
/pathology/prevention control MH -
Sulfonamides/therapeutic use MH - Support, U.S.
Gov't, P.H.S. MH - Ultraviolet Rays

• PMID- 10434051
• TI - Quantitative alterations of hyaluronan and
  dermatan sulfate in the
• hairless mouse dorsal skin exposed to
  chronic UV irradiation.
• MH - Animals
• MH - Chondroitin ABC Lyase
• MH - Collagenases
• MH - Deoxyribonucleases, Type II Site-Specific
• MH - Dermatan Sulfate/radiation effects
• MH - Disaccharides/analysis
• MH - Female
• MH - Histological Techniques
• MH - Hyaluronic Acid/radiation effects
• MH - Mice
• MH - Mice, Inbred HRS
• MH - Skin/chemistry/pathology/ radiation
  effects
• MH - Swine
• MH - Ultraviolet Rays

Analytical, Diagnostic and Therapeutic Techniques
and Equipment
Investigative Techniques
Clinical Laboratory Techniques
Histological Techniques
Cytological Techniques
Histocytochemistry
Immunohistochemistry
11
MeSH-based Representations
MeSH-based Implicit Link
Hypothesis network integration allows to link to
new information.
PMID- 10506108 TI - Reduction of UV-induced skin
tumors in hairless mice by selective COX-2
inhibition. MH - Animals MH - Carcinoma,
Squamous Cell/enzymology/pathology/prevention
control MH - Cell Division MH - Cyclooxygenase
Inhibitors/therapeutic use MH - Female MH -
Immunohistochemistry MH - Isoenzymes/drug
effects/metabolism MH - Mice MH - Mice, Inbred
HRS MH - Neoplasms, Radiation-Induced/enzymology/
pathology/prevention control MH -
Prostaglandin-Endoperoxide Synthase/drug
effects/metabolism MH - Skin Neoplasms/enzymology
/pathology/prevention control MH -
Sulfonamides/therapeutic use MH - Support, U.S.
Gov't, P.H.S. MH - Ultraviolet Rays

• PMID- 10434051
• TI - Quantitative alterations of hyaluronan and
  dermatan sulfate in the
• hairless mouse dorsal skin exposed to
  chronic UV irradiation.
• MH - Animals
• MH - Chondroitin ABC Lyase
• MH - Collagenases
• MH - Deoxyribonucleases, Type II Site-Specific
• MH - Dermatan Sulfate/radiation effects
• MH - Disaccharides/analysis
• MH - Female
• MH - Histological Techniques
• MH - Hyaluronic Acid/radiation effects
• MH - Mice
• MH - Mice, Inbred HRS
• MH - Skin/chemistry/pathology/ radiation
  effects
• MH - Swine
• MH - Ultraviolet Rays

12
Network-based Similarity Measures (term-term)
Edge-based vs. Node-based
Analytical, Diagnostic and Therapeutic Techniques
and Equipment
Investigative Techniques
Clinical Laboratory Techniques
Histological Techniques
Cytological Techniques
Histocytochemistry
Immunohistochemistry
13
Network-based Similarity Measures (term-term)
Edge-based vs. Node-based
Edge-based
Rada et al. (1989)
Analytical, Diagnostic and Therapeutic Techniques
and Equipment
Histological Techniques
2
Investigative Techniques
Cytological Techniques
Clinical Laboratory Techniques
Histological Techniques
Immunohistochemistry
Histological Techniques
Cytological Techniques
Histocytochemistry
Immunohistochemistry
14
Network-based Similarity Measures (term-term)
Edge-based vs. Node-based
Edge-based
Rada et al. (1989)
Analytical, Diagnostic and Therapeutic Techniques
and Equipment
Histological Techniques
2
Investigative Techniques
Cytological Techniques
Clinical Laboratory Techniques
Histological Techniques
4
Immunohistochemistry
Histological Techniques
Cytological Techniques
Histocytochemistry
Immunohistochemistry
15
Network-based Similarity Measures (term-term)
Edge-based vs. Node-based
Edge-based
Node-based
Resnik (1994)
Rada et al. (1989)
Analytical, Diagnostic and Therapeutic Techniques
and Equipment
Node Information Content Calculation - log (
P(node) )
85
Histological Techniques
2
Investigative Techniques
70
Cytological Techniques
Collection Frequencies
Clinical Laboratory Techniques
Histological Techniques
61
4
Immunohistochemistry
Histological Techniques
Cytological Techniques
35
23
Histocytochemistry
19
Immunohistochemistry
14
16
Network-based Similarity Measures (term-term)
Edge-based vs. Node-based
Edge-based
Node-based
Rada et al. (1989)
Resnik (1994)
Analytical, Diagnostic and Therapeutic Techniques
and Equipment
Node Information Content Calculation - log (
P(node) )
1
Histological Techniques
2
Investigative Techniques
0.82
Cytological Techniques
Relative Frequencies
Clinical Laboratory Techniques
0.72
Histological Techniques
4
Immunohistochemistry
Histological Techniques
Cytological Techniques
0.41
0.27
Histocytochemistry
0.22
Immunohistochemistry
0.16
17
Network-based Similarity Measures (term-term)
Edge-based vs. Node-based
Edge-based
Node-based
Resnik (1994)
Rada et al. (1989)
Analytical, Diagnostic and Therapeutic Techniques
and Equipment
Node Information Content Calculation - log (
P(node) )
0
Histological Techniques
2
Investigative Techniques
0.09
Cytological Techniques
Information Contents
Clinical Laboratory Techniques
0.14
Histological Techniques
4
Immunohistochemistry
Histological Techniques
Cytological Techniques
0.39
0.58
Histocytochemistry
0.66
Immunohistochemistry
0.80
18
Network-based Similarity Measures (term-term)
Edge-based vs. Node-based
Edge-based
Node-based
Rada et al. (1989)
Resnik (1994)
Analytical, Diagnostic and Therapeutic Techniques
and Equipment
Node Information Content Calculation - log (
P(node) )
0
Histological Techniques
2
Investigative Techniques
0.09
Cytological Techniques
Lowest Common Ancestor
Clinical Laboratory Techniques
Histological Techniques
0.14
4
Histological Techniques
Immunohistochemistry
Histological Techniques
Cytological Techniques
0.14
0.39
0.58
Cytological Techniques
Histocytochemistry
0.66
Histological Techniques
0.14
Immunohistochemistry
Immunohistochemistry
0.80
19
Network-based Similarity Measures (term-term)
Edge-based vs. Node-based

• Network Depth Scaling.
• Network Density integration.

20
Network-based Similarity Measures (term-term)
Depth Scaling
Anatomy
Cells
Respiratory System
Cellular Structures
Blood Cells
Respiratory Mucosa
Trachea
Intracellular Space
Leukocytes
Cell Surface Extensions
Cytoplasm
Leukocytes, Mononuclear
Cytoplasmic Structures
Lymphocytes
Monocytes
Organelles
Cytoskeleton
Microtubule-Organizing center
Intermediate Filaments
Microfilaments
Microtubules
Neurofibrils
Centrosome
Centrioles
21
Network-based Similarity Measures (term-term)
Density Integration
Anatomy
Cells
Respiratory System
Cellular Structures
Blood Cells
Respiratory Mucosa
Trachea
Intracellular Space
Leukocytes
Cell Surface Extensions
Cytoplasm
Leukocytes, Mononuclear
Cytoplasmic Structures
Lymphocytes
Monocytes
Organelles
Cytoskeleton
Microtubule-Organizing center
Intermediate Filaments
Microfilaments
Microtubules
Neurofibrils
Centrosome
Centrioles
22
Our Work
MeSH Link Generation with the Generalized Cosine
Similarity Measure (GCSM) (Ganesan et al, 2003).

• The Cosine Similarity Measure (CSM)

• The Generalized Cosine Similarity Measure (GCSM)

23
GCSM Network Integration
Term-term Similarity Measure
PMID- 10506108 TI - Reduction of UV-induced skin
tumors in hairless mice by selective COX-2
inhibition. MH - Animals MH - Carcinoma,
Squamous Cell/enzymology/pathology/prevention
control MH - Cell Division MH - Cyclooxygenase
Inhibitors/therapeutic use MH - Female MH -
Immunohistochemistry MH - Isoenzymes/drug
effects/metabolism MH - Mice MH - Mice, Inbred
HRS MH - Neoplasms, Radiation-Induced/enzymology/
pathology/prevention control MH -
Prostaglandin-Endoperoxide Synthase/drug
effects/metabolism MH - Skin Neoplasms/enzymology
/pathology/prevention control MH -
Sulfonamides/therapeutic use MH - Support, U.S.
Gov't, P.H.S. MH - Ultraviolet Rays

• PMID- 10434051
• TI - Quantitative alterations of hyaluronan and
  dermatan sulfate in the
• hairless mouse dorsal skin exposed to
  chronic UV irradiation.
• MH - Animals
• MH - Chondroitin ABC Lyase
• MH - Collagenases
• MH - Deoxyribonucleases, Type II Site-Specific
• MH - Dermatan Sulfate/radiation effects
• MH - Disaccharides/analysis
• MH - Female
• MH - Histological Techniques
• MH - Hyaluronic Acid/radiation effects
• MH - Mice
• MH - Mice, Inbred HRS
• MH - Skin/chemistry/pathology/ radiation
  effects
• MH - Swine
• MH - Ultraviolet Rays

24
GCSM Network Integration
Term-term Similarity Measure
PMID- 10506108 TI - Reduction of UV-induced skin
tumors in hairless mice by selective COX-2
inhibition. MH - Animals MH - Carcinoma,
Squamous Cell/enzymology/pathology/prevention
control MH - Cell Division MH - Cyclooxygenase
Inhibitors/therapeutic use MH - Female MH -
Immunohistochemistry MH - Isoenzymes/drug
effects/metabolism MH - Mice MH - Mice, Inbred
HRS MH - Neoplasms, Radiation-Induced/enzymology/
pathology/prevention control MH -
Prostaglandin-Endoperoxide Synthase/drug
effects/metabolism MH - Skin Neoplasms/enzymology
/pathology/prevention control MH -
Sulfonamides/therapeutic use MH - Support, U.S.
Gov't, P.H.S. MH - Ultraviolet Rays

• PMID- 10434051
• TI - Quantitative alterations of hyaluronan and
  dermatan sulfate in the
• hairless mouse dorsal skin exposed to
  chronic UV irradiation.
• MH - Animals
• MH - Chondroitin ABC Lyase
• MH - Collagenases
• MH - Deoxyribonucleases, Type II Site-Specific
• MH - Dermatan Sulfate/radiation effects
• MH - Disaccharides/analysis
• MH - Female
• MH - Histological Techniques
• MH - Hyaluronic Acid/radiation effects
• MH - Mice
• MH - Mice, Inbred HRS
• MH - Skin/chemistry/pathology/ radiation
  effects
• MH - Swine
• MH - Ultraviolet Rays

Organisms
Animals
25
GCSM Network Integration
Edge-based term-term Similarity Measure
PMID- 10506108 TI - Reduction of UV-induced skin
tumors in hairless mice by selective COX-2
inhibition. MH - Animals MH - Carcinoma,
Squamous Cell/enzymology/pathology/prevention
control MH - Cell Division MH - Cyclooxygenase
Inhibitors/therapeutic use MH - Female MH -
Immunohistochemistry MH - Isoenzymes/drug
effects/metabolism MH - Mice MH - Mice, Inbred
HRS MH - Neoplasms, Radiation-Induced/enzymology/
pathology/prevention control MH -
Prostaglandin-Endoperoxide Synthase/drug
effects/metabolism MH - Skin Neoplasms/enzymology
/pathology/prevention control MH -
Sulfonamides/therapeutic use MH - Support, U.S.
Gov't, P.H.S. MH - Ultraviolet Rays

• PMID- 10434051
• TI - Quantitative alterations of hyaluronan and
  dermatan sulfate in the
• hairless mouse dorsal skin exposed to
  chronic UV irradiation.
• MH - Animals
• MH - Chondroitin ABC Lyase
• MH - Collagenases
• MH - Deoxyribonucleases, Type II Site-Specific
• MH - Dermatan Sulfate/radiation effects
• MH - Disaccharides/analysis
• MH - Female
• MH - Histological Techniques
• MH - Hyaluronic Acid/radiation effects
• MH - Mice
• MH - Mice, Inbred HRS
• MH - Skin/chemistry/pathology/ radiation
  effects
• MH - Swine
• MH - Ultraviolet Rays

Analytical, Diagnostic and Therapeutic Techniques
and Equipment
Investigative Techniques
Clinical Laboratory Techniques
Histological Techniques
Cytological Techniques
Histocytochemistry
Immunohistochemistry
26
GCSM Network Integration
Depth Scaling
Anatomy
Cells
Respiratory System
Cellular Structures
Blood Cells
Respiratory Mucosa
Trachea
Intracellular Space
Leukocytes
Cell Surface Extensions
Leukocytes, Mononuclear
Cytoplasm
Cytoplasmic Structures
Lymphocytes
Monocytes
Organelles
Cytoskeleton
27
GCSM Network Integration
Density Integration
Anatomy
Cells
Respiratory System
Cellular Structures
Blood Cells
Respiratory Mucosa
Germ Cells
Neurons
Stem Cells
Trachea
Intracellular Space
Leukocytes
Cell Surface Extensions
Leukocytes, Mononuclear
Cytoplasm
Cytoplasmic Structures
Lymphocytes
Monocytes
Organelles
Cytoskeleton
28
GCSM Evaluation
TrecGen04 Collection
Text REtrieval Conference (TREC)
Genomics Track
TrecGen04 Collection - Subset of the MEDLINE
database (about 4.5 million biomedical
abstracts) - 50 topics - Judged documents for
each topic (about 49,000 in total, range of
500-1500 per topic)
29
GCSM Evaluation
Topic-specific Judged Documents
Relevant Non-relevant
Closely associated documents tend to be relevant
to the same request.
van Rijsbergen 1979
30
GCSM Evaluation
Partitional Clustering with CSM and GCSM
High-document-similarity cluster
Low-document-similarity cluster
Closely associated documents tend to be relevant
to the same request.
van Rijsbergen 1979
31
GCSM Evaluation
Partitional Clustering with CSM and GCSM
High-document-similarity cluster
Low-document-similarity cluster
Work and results presented at CORIA 2005
32
GCSM Evaluation
Averages over 50 topics
Initial 0.17
GCSM 0.22
CSM 0.28
33
GCSM Evaluation
Averages over 50 topics
GCSM 0.75
CSM 0.77
34
GCSM Evaluation
Averages over 50 topics
Initial 0.17
GCSM 0.22
CSM 0.28
35
GCSM Evaluation
Multiples Occurrences Problem (Network Density)
GCSM0.92
Doc 2
Doc 1
Anatomy
Anatomy
Cells
Cells
Cellular Structures
Cellular Structures
Intracellular Space
Intracellular Space
Cytoplasm
Cytoplasm
Cytoplasmic Structures
Cytoplasmic Structures
Cytoskeleton
Cytoskeleton
Microtubule-Organizing center
Intermediate Filaments
Microfilaments
Microtubules
Neurofibrils
36
GCSM Evaluation
Multiples Occurrences Problem (network density)
GCSM0.82
Doc 1
Doc 2
Anatomy
Anatomy
Cells
Cells
Cellular Structures
Cellular Structures
Blood Cells
Intracellular Space
Intracellular Space
Leukocytes
Cytoplasm
Cytoplasm
Leukocytes, Mononuclear
Cytoplasmic Structures
Cytoplasmic Structures
Lymphocytes
Cytoskeleton
Cytoskeleton
Neurofibrils
Neurofibrils
37
Related Work
Network-based Similarity Measures
Inter-term
Node-based
Edge-based
Rada et al (1989) Lee Kim (1993) Sussna
(1993) Wu Palmer (1994) Hirst St-Onge (1998)
Resnik (1995) Leacock Chodorow
(1994) Richardson Smeaton (1995) Jiang
Conrath (1997)
Inter-document
Ganesan et al (2003)
38
Network-based Similarity Measures
Past Evaluations
Our Focus
WordNet -Broad coverage of English (the noun
portion has over 107,400 noun senses) -Cognitive
approach -9 types of relationships
MeSH, GO -Domain specific. -Indexing, retrieval,
unification approaches. -Few relationship types
39
Network-based Similarity Measures
Past Evaluations
Our Focus
Term-term Similarities
Inter-document Similarities
Histological Techniques
Cytological Techniques
MH - Animals MH - Carcinoma, Squamous
Cell/enzymology/pathology/prevention
control MH - Cell Division MH - Cyclooxygenase
Inhibitors/therapeutic use MH - Female MH -
Immunohistochemistry MH - Isoenzymes/drug
effects/metabolism MH - Mice MH - Mice, Inbred
HRS MH - Neoplasms, Radiation-Induced/enzymology/
pathology/prevention control MH -
Prostaglandin-Endoperoxide Synthase/drug
effects/metabolism MH - Skin Neoplasms/enzymology
/pathology/prevention control MH -
Sulfonamides/therapeutic use MH - Support, U.S.
Gov't, P.H.S. MH - Ultraviolet Rays
MH - Animals MH - Chondroitin ABC Lyase MH -
Collagenases MH - Deoxyribonucleases, Type II
Site-Specific MH - Dermatan Sulfate/radiation
effects MH - Disaccharides/analysis MH -
Female MH - Histological Techniques MH -
Hyaluronic Acid/radiation effects MH - Mice MH
- Mice, Inbred HRS MH - Skin/chemistry/pathology/
radiation effects MH - Swine MH -
Ultraviolet Rays
40
Related Work
Genomic Information Retrieval
Ontrup et al. (2003) Srinivasan (1996) Shin et
al. (2004)
Network-based Similarity Measures
Inter-term
Node-based
Edge-based
Rada et al. (1989) Lee Kim (1993) Sussna
(1993) Wu Palmer (1994) Hirst St-Onge (1998)
Resnik (1995) Leacock Chodorow
(1994) Richardson Smeaton (1995) Jiang
Conrath (1997)
Inter-document
Ganesan et al. (2003)
41
Related Work
Genomic Information Retrieval
Ontrup et al. (2003) Srinivasan (1996) Shin et
al. (2004)
Network-based Similarity Measures
Inter-term
Node-based
Edge-based
Rada et al. (1989) Lee Kim (1993) Sussna
(1993) Wu Palmer (1994) Hirst St-Onge (1998)
Resnik (1995) Leacock Chodorow
(1994) Richardson Smeaton (1995) Jiang
Conrath (1997)
Genomic Information Discovery
Struble and Dharmanolla (2004) Srinivasan
(2004). Shatkay et al. (2000) Jenssen et al.
(2001)
Inter-document
Ganesan et al. (2003)
42
PhD Hypothesis

• The Big Hypothesis Genomic ontologies and
  network-based similarity measures can help us
  create implicit links in Genomic Information to
  improve Information Retrieval and Discovery.
• Our Hypothesis Genomic ontology networks (MeSH
  and GO) can be used to generate new links between
  biomedical research articles and genomic
  databases records.

43
PhD Plan
MeSH-based Link Creation
Go-based Link Creation
ctcggagggc accgcccacc gggg

• Experiments
• Document MeSH Representations.
• Text-based vs. MeSH-based.
• MeSH-based with/without network.
• Edge-based vs. Information-based.
• Impact of Depth Scaling and Density Integration.
• Impact of Multiple Occurrences Problem.

cagccgaggc ctcggagggc accg

• Experiments
• Sequence-based vs. GO-based.
• GO-based with/without network.
• Edge-based vs. Information-based.
• Impact of Depth Scaling and Density Integration.
• Impact of Multiple Occurrences Problem.

• Evaluation
• Partitional Clustering.
• Hierarchical Agglomerative Clustering.
• Relevance Feedback.

Evaluation and Dataset Still under
investigation Comparison with literature? Compari
son with Gene-interaction data?
Dataset TrecGen04 (TREC Genomics Track)
44
Thank you for attending