Generating Biomedical Hypotheses Using Semantic Web Technologies

Generating (and Testing) Biomedical Hypotheses
Using Semantic Web Technologies

Michel Dumontier, Ph.D.
Associate Professor of Medicine (Biomedical Informatics)
Stanford University

1

@micheldumontier::AAAI-FSW:Nov 16, 2013

2


Science, it works ******

3


We use the biomedical literature to find out what we
believe to be true

4


we need to access to the most recent biomedical data
(problems: access, format, identifiers & linking)

5


we also need access to the most effective software
to analyze, predict and evaluate
(problems: OS, versioning, input/output formats)

6


we build support for our hypotheses by constructing
and (partly) reusing fairly sophisticated workflows

7


Wouldn’t it be great if we could just find the evidence
required to support or dispute a scientific hypothesis using
the most up-to-date and relevant data, tools and scientific
knowledge?
8


madness!
1. Build a massive network of interconnected
data and software using web standards
2. Develop methods to generate and test
hypotheses across these data
1. tactical formalization
2. uncovering associations
3. evidence gathering

3. Contribute back to the global knowledge
graph
9


The Semantic Web
is the new global web of knowledge
standards for publishing, sharing and querying
facts, expert knowledge and services
scalable approach for the discovery
of independently formulated
and distributed knowledge

10


we’re building an incredible network of linked data

11 Linking Open Data cloud diagram, by Richard Cyganiak and Anja Jentzsch


Bio2RDF Linked Data Map

Bio2RDF Release 2: Improved coverage, interoperability and provenance of Life Science Linked Data . ESWC 2013.
12


At the heart of Linked Data for the Life Sciences

chemicals/drugs/formulations,
genomes/genes/proteins, domains
Interactions, complexes & pathways
animal models and phenotypes
Disease, genetic markers,
treatments
Terminologies & publications

• Free and open source
• Uses Semantic Web standards
• Release 2 (Jan 2013): 1B+ interlinked statements from 19 conventional and high
value datasets
• Includes provenance, statistics
• Partnerships with EBI, NCBI, DBCLS, NCBO, OpenPHACTS, and commercial tool
providers

13

A Callahan, J Cruz-Toledo, M Dumontier. Querying Bio2RDF Linked Open Data with a Global Schema.
2013. Journal of Biomedical Semantics. 4(Suppl 1):S1.

Resource Description Framework
• It’s a language to represent knowledge
– Logic-based formalism -> automated reasoning
– graph-like properties -> data analysis

• Good for
– Describing in terms of type, attributes, relations
– Integrating data from different sources
– Sharing the data (W3C standard)
– Reuse what is available, develop what you’re
missing.
14

Dumontier::FDA:Nov 15,2013

drugbank_vocabulary:Drug

rdf:type
rdfs:label
drugbank:DB00586

Diclofenac [drugbank:DB00586]

drugbank_vocabulary:targets
drugbank_target:290

rdfs:label

Prostaglandin G/H synthase 2
[drugbank_target:290]

rdf:type
drugbank_vocabulary:Target
15


The linked data network expands
with every reference
DrugBank

drugbank_vocabulary:Drug

rdf:type
rdfs:label
drugbank:DB00586

diclofenac [drugbank:DB00586]

pharmgkb_vocabulary:xref
pharmgkb:PA449293

rdfs:label

PharmGKB

diclofenac [pharmgkb:PA449293]

pharmgkb_vocabulary:Drug
16


Federated Queries over Independent
SPARQL EndPoints
Get all protein catabolic processes (and more specific) from biomodels

SELECT ?go ?label count(distinct ?x)
WHERE {
service <http://bioportal.bio2rdf.org/sparql> {
?go rdfs:label ?label .
?go rdfs:subClassOf ?tgo
?tgo rdfs:label ?tlabel .
FILTER regex(?tlabel, "^protein catabolic process")
}
service <http://biomodels.bio2rdf.org/sparql> {
?x <http://bio2rdf.org/biopax_vocabulary:identical-to> ?go .
?x a <http://www.biopax.org/release/biopax-level3.owl#BiochemicalReaction> .
}
}

17


Bio2RDF: 1M+ SPARQL queries per month
UniProt: 6.4M queries per month
EBI: 3.5M queries (Oct 2013)

18


Directions
• research
– approaches for data publication
– methods for large scale data reduction
– methods for mining associations

• service
– coverage, quality and licensing
– better user interfaces
– sustainability of service
19


tactical formalization
a. discovery of drug and disease pathway associations
b. prediction of drug targets using phenotypes

20


ontology as a
strategy to formally
represent and
integrate knowledge

21


Have you heard of OWL?

22


Identification of
disease enriched pathways
def: A biological pathway is constituted by a
set of molecular components that undertake
some biological transformation to achieve a
stated objective
glycolysis : a pathway that converts glucose to
pyruvate

23


aberrant pathways
which pathways are associated with disease?
Which pathways can be perturbed by drugs?
disease
pathway

drug
24


Identification of
drug and disease enriched pathways
• Approach
– Integrate 3 datasets: DrugBank, PharmGKB and
CTD
– Integrate 7 terminologies: MeSH, ATC, ChEBI,
UMLS, SNOMED, ICD, DO
– Identify significant pathways via enrichment
analysis over the fully inferred knowledge base

• Formalized as an OWL-EL ontology
– 650,000+ classes, 3.2M subClassOf axioms, 75,000
equivalentClass axioms
25

Identifying aberrant pathways through integrated analysis of knowledge in pharmacogenomics.
Bioinformatics. 2012.

Top Level Classes
(disjointness)

pathway

drug

gene

disease
Reciprocal
Existentials

Class subsumption

mercaptopurine
[pharmgkb:PA450379]

purine-6-thiol
[CHEBI:2208]

Class Equivalence
mercaptopurine
[drugbank:DB01033]

mercaptopurine
[mesh:D015122]
mercaptopurine
[ATC:L01BB02]

drug
property chains

pathway

drug
disease

gene
26


Benefits: Enhanced Query Capability
– Use any mapped terminology to query a target resource.
– Use knowledge in target ontologies to formulate more
precise questions
• ask for drugs that are associated with diseases of the joint:
‘Chikungunya’ (do:0050012) is defined as a viral infectious disease
located in the ‘joint’ (fma:7490) and caused by a ‘Chikungunya
virus’ (taxon:37124).

– Learn relationships that are inferred by automated
reasoning.
• alcohol (ChEBI:30879) is associated with alcoholism (PA443309)
since alcoholism is directly associated with ethanol (CHEBI:16236)
• ‘parasitic infectious disease’ (do:0001398) retrieves 129 disease
associated drugs, 15 more than are directly associated.

27


Knowledge Discovery through Data
Integration and Enrichment Analysis
• OntoFunc: Tool to discover significant associations between sets of objects
and ontology categories. enrichment of attribute among a selected set of
input items as compared to a reference set. hypergeometric or the
binomial distribution, Fisher's exact test, or a chi-square test.
• We found 22,653 disease-pathway associations, where for each pathway
we find genes that are linked to disease.
– Mood disorder (do:3324) associated with Zidovudine Pathway
(pharmgkb:PA165859361). Zidovudine is for treating HIV/AIDS. Side
effects include fatigue, headache, myalgia, malaise and anorexia
• We found 13,826 pathway-chemical associations
– Clopidogrel (chebi:37941) associated with Endothelin signaling
pathway (pharmgkb:PA164728163). Endothelins are proteins that
constrict blood vessels and raise blood pressure. Clopidogrel inhibits
platelet aggregation and prolongs bleeding time.
28


PhenomeDrug
A computational approach to predict drug
targets, drug effects, and drug indications using
phenotypic information

Mouse model phenotypes provide information about human drug targets.
Hoehndorf R, Hiebert T, Hardy NW, Schofield PN, Gkoutos GV, Dumontier M.
Bioinformatics. 2013.
29


animal models provide insight for on target effects
• In the majority of 100 best selling drugs ($148B in US
alone), there is a direct correlation between knockout
phenotype and drug effect
• Gastroesophageal reflux
– Proton pump inhibitors (e.g Prilosec - $5B)
– KO of alpha or beta unit of H/K ATPase are achlorhydric (normal
pH) in stomach compared to wild type (pH 3-5) when exposed
to acidic solution

• Immunological Indications
– Anti-histamines (Claritin, Allegra, Zyrtec)
– KO of histamine H1 receptor leads to decreased responsiveness
of immune system
– Predicts on target effects : drowsiness, reduced anxiety
Zambrowicz and Sands. Nat Rev Drug Disc. 2003.
30


Identifying drug targets
from mouse knock-out phenotypes
Main idea: if a drug’s phenotypes matches the phenotypes of a
null model, this suggests that the drug is an inhibitor of the gene
phenotypes

similar
effects

non-functional
gene model

drug

ortholog

gene
31

inhibits
human gene

Terminological Interoperability
(we must compare apples with apples)
Mouse
Phenotypes

PhenomeNet

Mammalian
Phenotype
Ontology

PhenomeDrug

Drug effects
(mappings from UMLS to DO, NBO, MP)

Semantic Similarity
Given a drug effect profile D and a mouse model M, we
compute the semantic similarity as an information weighted
Jaccard metric.

The similarity measure used is non-symmetrical and
determines the amount of information about a drug effect
profile D that is covered by a set of mouse model
phenotypes M.

33


null models do well in predicting
targets of drug inhibitors
• 14,682 drugs; 7,255 mouse genotypes
• Validation against known and predicted inhibitor-target pairs
– 0.739 ROC AUC for human targets (DrugBank)
– 0.736 ROC AUC for mouse targets (STITCH)
– 0.705 ROC AUC for human targets (STITCH)

• diclofenac (STITCH:000003032)
– NSAID used to treat pain, osteoarthritis and rheumatoid arthritis
– Drug effects include liver inflammation (hepatitis), swelling of liver
(hepatomegaly), redness of skin (erythema)
– 49% explained by PPARg knockout
• peroxisome proliferator activated receptor gamma (PPARg) regulates metabolism,
proliferation, inflammation and differentiation,
• Diclofenac is a known inhibitor

– 46% explained by COX-2 knockout
• Diclofenac is a known inhibitor

Finding evidence to support or dispute a hypothesis takes
some digging around

35


FDA Use Case:
TKI non-QT Cardiotoxicity
• Tyrosine Kinase Inhibitors (TKI)
– Imatinib, Sorafenib, Sunitinib, Dasatinib, Nilotinib, Lapatinib
– Used to treat cancer
– Linked to cardiotoxicity.

• FDA launched drug safety program to detect toxicity
– Need to integrate data and ontologies (Abernethy, CPT 2011)
– Abernethy (2013) suggest using public data in genetics,
pharmacology, toxicology, systems biology, to
predict/validate adverse events

• What evidence could we gather to give credence that
TKI’s causes non-QT cardiotoxicity?
36


HyQue
• The goal of HyQue is retrieve and evaluate evidence
that supports/disputes a hypothesis
– hypotheses are described as a set of events
• e.g. binding, inhibition, phenotypic effect

– events are associated with types of evidence
• a query is written to retrieve data
• a weight is assigned to provide significance

• Hypotheses are written by people who seek answers
• data retrieval rules are written by people who know the
data and how it should be interpreted 
1. HyQue: Evaluating hypotheses using Semantic Web technologies. J Biomed Semantics. 2011 May 17;2 Suppl 2:S3.
2. Evaluating scientific hypotheses using the SPARQL Inferencing Notation. Extended Semantic Web Conference (ESWC
2012). Heraklion, Crete. May 27-31, 2012.
37


HyQue: A Semantic Web Application

Hypothesis

Evaluation

Data

Ontologies

Software

@micheldumontier::AAAI-

38

SPIN functions
to evaluate TKI Cardiotoxicity
• Effects [r3.sider, lit]
• Does the drug have known cardiotoxic effects?

• Models [r2.mgi]
• Are there cardiotoxic phenotypes in null mouse models of target genes?

• Assays [ebi.chembl]
• Is hERG inhibited?

• Is the drug TUNEL positive?

• Targets [r2.drugbank, lit]
• Does the drug inhibit cardiotoxic-related proteins?
• RAF1, PDFGFR, VEGFR, AMPK

39


40


http://bio2rdf.org/drugbank:DB01268
41


42


43


In Summary
• This talk was about making sense of the
structured data we already have
• RDF-based Linked Open Data acts as a substrate
for query answering and task-based formalization
in OWL
• Discovery through the generation of testable
hypotheses in the target domain.
• The next big challenge lies in capturing the
output of computational experiments just as
much as extracting/formalizing user-contributed
knowledge .
44


Acknowledgements
Bio2RDF Release 2:
Allison Callahan, Jose Cruz-Toledo, Peter Ansell

Aberrant Pathways: Robert Hoehndorf, Georgios Gkoutos
PhenomeDrug: Tanya Hiebert, Robert Hoehndorf, Georgios
Gkoutos, Paul Schofield
HyQue: Alison Callahan, Nigam Shah

45


dumontierlab.com
michel.dumontier@stanford.edu
Website: http://dumontierlab.com
Presentations: http://slideshare.com/micheldumontier

46


Generating Biomedical Hypotheses Using Semantic Web Technologies

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Generating Biomedical Hypotheses Using Semantic Web Technologies (20)

More from Michel Dumontier (20)

Recently uploaded (20)

Generating Biomedical Hypotheses Using Semantic Web Technologies

Editor's Notes