ONSTR: Ontology for Newborn Screening Follow-up and Translational Research 2013 Joint Meeting of the Newborn Screening and Genetic Testing Symposium th Thursday, May 9 Atlanta Prabhu Shankar MD, MS Snežana Nikoli ć , MA, Sivaram Arabandi, MD, MS Shamakant Navathe Kunal Malhotra Rani H. Singh PhD, RD
Objective • NBS and f NBS and follow-up w llow-up workflows and da ows and data comple ta complexity xity • Ontology and Semantic W Ontology and Semantic Web tec b technologies hnologies • ONSTR ONSTR llow-up Da Data Inte ta Integration ion • Ne Newborn Scr wborn Screening F ening Follow-up Colla Collabor borativ tive (NBSD (NBSDC) • Semantic W Semantic Web tec b technology success stories in nology success stories in healthcar healthcare – – if time permits! time permits!
External Computational Support! “OMICS ”
In Summary NBS Da NBS Data is: ta is: • Geog Geographically distrib phically distribut uted ed (data silos!) • Inter Intersects c ects clinical as w inical as well as many ll as many biomedical domains biomedical domains, e.g., biochemistry, pathways, metabolomics, genomics, proteomics, pharmacogenomics • Various f rious forma rmats – s – str tructur uctural, sc l, schema hematic tic and semantic v and semantic varia riability bility • Of Of r rare diseases! diseases! 5
Semantic Web Technologies …technology stack to support a “Web of data” "The Semantic Web pr provides a ides a common common fr frame amework k that allows data to be shared and reused across application, enterprise, and community boundaries.” Wo World W Wide We Web C Consortium ( (W3C)
Triples: Resource Description Framework (RDF) Resource Description Framework (RDF) PREDICATE SUBJECT OBJECT Relationships/Properties is_a Amino Acid Phenylalanine Infers! material_basis Phenylketonuria PAH Mutation Directed, edge-labeled graph is_a Inherited Phenylketonuria Metabolic Disease Simple, Dynami Simple Dynamic, Extensib Extensible, Inter Interoper perable RDF Sc RDF Schema (RDFS), W hema (RDFS), Web Ontology Langu b Ontology Language ‘Ontolo ‘Ontology gy’ ’
What is ontology? 1. A branch of philosophy, studying categories and types of beings existing in the universe. 2. In Informatics, explicit f plicit formal specif rmal specifica cations tions of the terms in the domain and relationships among them. - Consensus based - Associated with documentation and definitions - Expr Expressed essed in f in formal logic rmal logic to support automated reasoning - Interpretable by humans and computers
Semantic Methods and Characteristics Deanna Pennington, LTER DataBits, Spring 2006 Definition Synonyms Classification Properties Other (isa) (has) relations Method Keywords Dictionary X Controlled (X) X vocabulary Thesaurus X X Taxonomy (X) X X (X) X Parrot is a Parrot has bird a beak Ontology X X X X X You can search by a term’s properties
Relational? Periventricular Periventricular Genetic Disease White Matter Developmental White matter White matter has_MRI_finding Neurotoxin Corpus Callosum is_a is_a PAH Mutation Increased Blood Phe levels material_basis is_a PKU Fair skin & hair has_finding Autosomal eczema Recessive is_a Genetic Disease Genetic Disease seizures cognitive has_diagnostic_ Inherited Metabolic Inherited Metabolic procedure performance Disorder has_drug Measured_by Plasma A Acids Sapropterin Sapropterin has_diet_ Gene Analysis Large Neutral analyses Amino Acids Phe Intake/Day Phe Intake/Day is_a (hierarchical) Associative relations Schema-flexible!
Ontology Applications Naming Information “Things” Integration Natural Knowledgebase Language Ontologies e.g., Foundational Model of Processing Anatomy(FMA) (NLP) Data Analysis Data Exchange e.g., e.g., Gene Ontology(GO) Biological Pathway Exchange(BioPAX)
Natural Language Understanding! http://sc6.blogspot.com/2011/02/cartoon-of-week_20.html
What is ONSTR? An applica plication ontology ion ontology representing the processes, entities and knowledge in the Newborn Screening and follow-up system (Domain): Newborn screening Dried Blood Spot (NDBS) covering • Inherited Metabolic Diseases (IMDs IMDs). Genetic basis Genetic basis of IMDs. • Positive tested cases follow-up practice including: • medical/clinical medical/c linical conf confirma irmato tory testing (bioc ry testing (biochemical emical and and molecular). molecular). Medical Medical and n and nutritional tritional tr trea eatment tment (dietary analysis monitoring) • Outcomes, e.g., phy Outcomes physical sical and cognitive growth and owth and • de development e lopment evalua aluation. tion. Research related to IMDs and NBS. •
Why are we building ONSTR? To provide basis for standar standardiza ization of ion of • da data ta annotation in NBS domain. To provide knowledg knowledge base e base for • integrating, aggregating and reasoning over data collected from different NBS sources. To de develop lop tools ools for knowledge and data • sharing to be used by greater IMD/NBS community.
Not Alone…… Open Biomedical Ontologies (OBO) Foundry principles and framework.
ONSTR building process 1. Use case Def 1. Use case Definition nition ‘of all the diagnosis confirmed patients who were new born screening positive, between 2005-2010 2005-2010, matching age and matching mutation (R408W), did good nutritional ma utritional mana nagement ment VS K VS Kuvan + Nutritional Nutritional mana management ment had better outcom outcome with regards to MRI Whi MRI White ma e matter c tter chang anges s at five years?’. 2. Identif 2. Identifica cation of tion of k key entities and r y entities and rela lationships tionships holding betw holding between these entities een these entities Methodology: Methodology: - Top Down and Bottom Up - Survey of relevant literature - Identifying the common data elements (CDEs) - Follow OBO Foundry best practices
Top Down and Bottom Up Common Data Elements (CDEs)
All Common Data Elements (CDEs)
Modeling with Relations Blood Collection Procedure Hearing Annotated Testing has_Participant Filter Paper Is_a BFO: Process Is_a has_Participant Aggregate BFO: Pulse Is_a Process Heel Prick Blood Finger Prick Oximetry State Specific Specimen Testing Procedure Collection Blood Collection II Screening Container Procedure Procedure part_of Is_a Is_a part_of BFO: part_of outPut_of outPut_of Process New Born Health Screening Specimen Genetic Is_a Screening (NBS) Testing Dried Blood part_of Procedure Spot (DBS) Is_a part_of part_of Is_a Biochemical preparedFrom Testing Blood BFO: Specimen Process hasTestSpecimen part_of Biomarker Concentration collected_from hasOutput Dried Blood Blood Tyrosine Newborn BFO: Spot Elute Concentration Is_a Process Is_a Blood isTestSpeci Biomarker Phenylalanine menFor Assay Concentration
ONSTR building process contd. 3. Ontology coding - ONSTR is formally encoded as a RDF/XML seriali RDF/XML serialization of ion of OWL2 (W3C semantic W L2 (W3C semantic Web b standar standards) s) 4. Ontology integration - Mappings between ONSTR and other relevant ontologies/vocabularies (Future work). 5. Ontology evaluation - I n progress, concomitant with ONSTR development. 6. Ontology documentation - Available on the ONSTR project page: http://code.google.com/p/onstr/source/docs )
ONSTR graph and Logical Definitions
ONSTR Statistics Total n tal number of mber of c classes: 1842 lasses: 1842 ONSTR na ONSTR nativ tive c classes: 1100 asses: 1100 Impor Imported c ed classes: 742 asses: 742
BioPortal http://bioportal.bioontology.org/ontologies/49978 National Center for Biomedical Ontology (NCBO), Stanford University
Ne Newborn Scr wborn Screening F eening Follow-up Da llow-up Data ta Inte Integration Colla ion Collabor borativ tive (NBSDC) (NBSDC)
Acknowledgements: Funded by: • 2009-10 2009-10 HSI HSI Seed eed Gr Grant ant, a Clinical Outcomes Research and Public Health (CORPH) Pilot Grants Program, jointly supported by Geor Georgia gia Tec ech and nd Childr Children’ en’s Healthcar Healthcare of Atlanta tlanta and • The Southeast Southeast NBS BS & Genetics enetics Colla ollabor borativ tive (SER (SERC) C) Grant from the Maternal and Child Health Bureau, HRSA Gr HRSA Grant ant U22MC10979. 22MC10979. Special thanks to: • Dr Dr. Bar Barry ry Smith Smith, National Center for Ontological Research (NCOR NCOR), University @ Buffalo, Buffalo.
Thank You Questions: PRSHANK@emory.edu 27
Semantic technologies in action…. Cross‐Species Biomarkers Reducing Animal Testing Result: Semantic integration (large animals to small animals to cell culture) to discover cross ‐ species biomarkers applicable to human adverse events and diseases Cour Courtes tesy: Eric rich Gombocz, VP Gombocz, VP & & CSO CSO, IO Inf IO Informa rmatics ics, Inc Inc.
Semantic technologies in action…. Combination Treatment Effectiveness in Prostate Cancer Result: effectiveness comparison of different combination treatments based on multi ‐ platform genomic and proteomic marker profiles and patient match
Recommend
More recommend
