open phacts webinar june 2016 - data2discovery
TRANSCRIPT
DATA2DISCOVERY
OpenPHACTS: Maximizing Impact for
Pharmaceutical Applications
DAVIDWILD,PhD.
CEOData2DiscoveryInc
AssociateProfessorandDirectorofDataScience,IndianaUniversity
http://d2discovery.com-http://[email protected]
DATASCIENCEDATA2DISCOVERY
1
DATA2DISCOVERY
Data2DiscoveryInc
• Findinginsightsfromproprietary,publicand
commercialdatatogetherinwaysnevertriedbefore
• Linkingdatafrommoleculartohuman
• KeyexpertiseinusingOpenPHACTSdata
• Partneringwithcustomerstoimplementthisatscale
• Phenotypicdrugdiscoveryisabeachheadapplication
• http://d2discovery.com
2
DATA2DISCOVERY
Data2DiscoveryApplicationAreas
PharmaceuticalResearchHealthcareDelivery
PublicHealth
PrecisionMedicine
PopulationHealth
RealWorldEvidence
Combinationtherapy
OutcomeAnalytics
NudgeEconomics
QuantifiedSelf
Performance-basedPricing
RiskAdjustment
PhenotypicDrugDiscovery
AdverseEventPrediction
3
DATA2DISCOVERY
ITInfrastructureRe-Imagined
4
WhateffectwillinhibitingPPAR-γhaveonHER+breastcancer
tissue?
InteractionInfrastructure(II)Custominterfacesandapplicationsbuiltusingreusablecomponents
ComputationInfrastructure(CI)Fast,inmemorycomputationmodulesthatworkonheterogeneousdatasub-graphs
LinkedDataEcosystem(LDE)DataAPIs,MicroTranslators,LiveTripleStores,Ontologies&Standards
Whatadverseeffectscould
Cymbaltacauseinpatientswith
Chron’s?
DATA2DISCOVERY
Casestudy–EliLilly
• Linkeddataecosystem:DevelopedliveData
Translatorthatmapskeyinternalassaydata
intoOpenPHACTStriplestore,annotatedby
publicontologies(e.g.BAO)
• Computation/Interactioninfrastructure:
Contextualinquiryinhighimpactend-user
applicationsacrosstheorganization
5
DATA2DISCOVERY
Method
Target UniProtChEMBLAssay
InternalAssayData
ChEMBLCompd
SemanticSchema
Assay
ResultType
ChEMBLCellLine
GeneCellLine
MappingInternalandExternalAssayData
6
DATA2DISCOVERY
PhenotypicDrugDiscovery
• “Beachhead”application
• SubjectofstrongPharmainterest
(OpenPHACTSresearchathonSpring2015)
• Crossesthebigtwodomains
– Chemistry&biology(molecular)
– Studyofpeople(patient)
• Needtofindlinksacrossmanydatasets
– Preclinical,toxicology,clinicaltrials,post-market
7
DATA2DISCOVERY
P3–PredictivePhenotypicProfiler
• P3isaproductindevelopmentbyData2Discoverythatisdesignedto
addresscurrentgapsinmaximizingimpactofphenotypicassaydata.
Intendedapplicationsinclude:
– Targetdeconvolution
– Target-basedmechanismofactiondiscovery
– Identifyingsimilaritiesbetweenphenotypicassays
• TheseusecaseswereidentifiedinOpenPHACTSPhenotypicScreening
WorkshopinFebruary2015
• IdentifiesassociationsusingSEMAPTMassociationfindingtechnology
• LinksOpenPHACTSdataintootherkeydataresources
8
DATA2DISCOVERY
P3Architecture
• Componentizedarchitecture
– Permitsplug-and-playofdatasets
– Allowsnewapplicationstobequicklydeveloped
• Highlyscalable
– Usescloudtechnology,fastgraphdatabasesandApacheSPARK
• Securedeploymentswhereneeded
– Canbeusedasexternalsoftware-asservice
– Canbedeployedbehindfirewalls(Dockerimages)
9
DATA2DISCOVERY
WhattypesofdatacanP3use?• P3canlinkpublicandproprietarydatasourcesandcrosspre-
clinicalandclinicaldata.Exampletypesofdatainclude:
– Enzymaticassay(compound-gene/target)
– Phenotypicassay(compound-phenotype)
– Cellularassay(compound-cellline)
– Geneexpression(e.g.LINCS,L1000)
– Pathway(pathway-gene)
– Molecular-Phenotypelinks(compound,gene,pathway,celllineto
diseasestate,genotype,patientphenotype,adverseevent,electronic
medicalrecord,realworldevidenceetc)
10
DATA2DISCOVERY
SEMAP™-SemanticAssociationPrediction
• BasedonresearchatIndianaUniversity
• Predictsassociationbasedondatasubnetworksbetweenpointsof
interest
• Canpredictdrug-targetinteractionacrossthousandsofgenes
• Noscopeorbiasproblems
• ExtensiveexternalvalidationincludingcomparisonwithSEAandusefor
predictinggeneexpressionprofiling(seeChen,B.etal.,PLoS
ComputationalBiology,2012,8(7),e1002574)
• Demonstratedapplicationsindrugrepositioning,MOAdiscovery
• NowbeingusedinphenotypictargetdeconvolutionandMOAdiscovery
11
DATA2DISCOVERY
Example:TroglitazoneandPPARGAssociationscore:2385.9Associationsignificance:9.06x10-6=>missinglinkpredicted
12
DATA2DISCOVERY
P3PublicDataDemonstration• LinksseveralpublicdatasetsusingTBasasampleTA:
• ChEMBL/OpenPHACTS(enzymatic,cellularassays)
• NCATSPhenotypicDrugDiscoveryResource1
(phenotypicassays)
• Manualphenotypicassay–pathway/geneannotations
• TBdrug–gene(DrugBank,NCATS)
1https://ncats.nih.gov/expertise/preclinical/pd213
DATA2DISCOVERY
OpenPhenotypicDrugDiscoveryResource
EliLillyIT
https://ncats.nih.gov/expertise/preclinical/pd214
DATA2DISCOVERY
CurrentP3LinkedDataNetwork
15
DATA2DISCOVERY
P3PublicDataExample
16
DATA2DISCOVERY
DrugDiscovery Development&ClinicalTrials
PostMarket/Healthcare
Use
Product
ExampleData
OpenPHACTS
OPDDRToxCast
InternalCpd/Assay
InternalTrialsData
ClinicalTrials.gov
QuantifiedSelf/IOT
MedicareCMSEMR
DataRWEData
P3-Discovery P3-Clinical P3-PopHealth
TargetID
Poly-Pharmacology
MOADiscovery
ADME/Tox
AdverseEventPrediction
CandidateRepositioning
DrugRepositioning
PopulationAnalysis
DrugCombinations
PersonalizedMedicine
FAERS
Value/PerformancebasedPricing
P3-DevelopmentPlan
17
DATA2DISCOVERY
WhatcanData2Discoveryoffer?• Agiledevelopmentofhighimpactapplications
thatusesemanticlinkeddata
• Implementationoflinkeddataecosystem,
computationinfrastructureanduser
components
• KeyexpertiseinusingOpenPHACTSdata
• ContactDavidWild,[email protected]
18