Life Web Science 2013, ParisImproving Transparency and Reproducibilityof Clinical ResearchUsing Semantic TechnologiesSorou...
Can we make the Web ascientific research platformfrom hypothesis right through to publication
Focus on publishing citable snippetsof scientific knowledge using SemWeb standards
That’s a good start v.v. academic publishing, but...What about the rest of the scientific process?
PublicationDiscourseHypothesisExperimentInterpretation
Houston, we have a problem...
ContextMultiple recent surveys of high-throughput biologyreveal that upwards of 50% of published studiesare not reproducib...
ContextSimilar (if not worse!) in clinical studies- Begley & Ellis, Nature, 2012- Booth, Forbes, 2012- Huang & Gottardo, B...
Context“the most common errors are simple,the most simple errors are common”At least partially because theanalytical metho...
ContextThese errors pass peer reviewThe researcher is unaware of the errorThe process that led to the error is not recorde...
ContextDiscovery of such errors have resulted in retractionsand even shut-down clinical trials- Ioannidis, 2009
ContextIn March, 2012, the US Institute of Medicine ~said“Enough is enough!”
ContextInstitute of Medicine RecommendationsFor Conduct of High-Throughput Research:Evolution of Translational Omics Lesso...
How do we get there from here?
Our early attempts atsupporting clinical researchthrough SemWeb
MEASUREMENT UNITSProblem #1: Integrating Clinical Data
Units must be expressedUnits must be harmonizedDon’t leave this up to the researcher(it’s fiddly, time-consuming, and erro...
NASA Mars Climate Orbiter
Oops!
QUDT: Quantities, Units, Dimensions and Types Ontologyconversion offset & conversion multiplierenable conversion between n...
OM: Ontology of Units of MeasurementUseful for inventing new units that are commonly used in clinical researchbut not (cur...
ID HEIGHT WEIGHT SBP CHOL HDL BMIGRSBPGRCHOLGRHDLGRpt1 1.82 177 128 227 55 0 0 1 0pt2 179 196 13.4 5.9 1.7 1 0 1 0Legacy c...
GOAL: get the clinical researcher“out of the loop” once the data is collectedComplete transparency in analysis withNO PEEK...
Extending the GALEN ontology with richer logicincluding measurement values and unitsmeasure:SystolicBloodPressure =galen:S...
Now Galen classes can be used to “carry” rich dataMove beyond use of ontologies for simple keyword curation(keyword hierar...
Now, what do we do with the unit-soup that is in our legacy dataset?
SADI Semantic Web Service for automated Unit conversion• Send it a dataset with mixed units• (optional) tell it the harmon...
Create additional ontological classesrepresenting clinical features of interest based on clinical guidelinesmeasure:HighRi...
SELECT ?record ?convertedvalue ?convertedunitFROM <./patient.rdf>WHERE {?record rdf:type measure:HighSystolicBloodPressure...
ASSESSING RISKProblem #1: Automatic Interpretation of Clinical Data
Framingham risk measurements:AgeGenderHeightWeightBody Mass Index(BMI)Systolic Blood Pressure(SBP)Diastolic Blood Pressure...
Measurements like BMI are derived from calculationsover other “core” measurementsAgain, we use SADI and semantics to achie...
Semantic Modeling of theAmerican Heart Association Risk gradesHighRiskBMI =PatientRecord and(sio:hasAttribute some(cardio:...
SHARE Query for High Risk SBP(SHARE is a SADI-enhanced SPARQL query engine)
Now the interesting question...
How does our automated risk evaluationcompare to a clinician’s expert risk classification?
True positive rate“at risk” %False positive rate“at risk”%SBP 100 0DBP 100 0CHOL 92.6 0HDL 100 56.5TG 100 8.5BMI 100 18.8L...
But... We encoded and were following the guidelines!We double-checked and our definitions were definitely correctHow could...
Visual inspection of data and guidelines showedin many cases the clinician had “tweaked” the guideline------------------AH...
Adjusting our OWL class definitions and re-running the analysisResulted in nearly 100% correspondence with the clinical re...
Reflect on this for a second... Because this is important!1. We automated data cleansing and analysis using Semantic Web S...
AHA:HighRiskCholesterolRecordPatientRecord and(sio:hasAttribute some(cardio:SerumCholesterolConcentration andsio:hasMeasur...
To do the “experiment” using AHL guidelinesSELECT ?patient ?riskWHERE {?patient rdf:type AHL: HighRiskCholesterolRecord .?...
To do the “experiment” using McManus guidelinesSELECT ?patient ?riskWHERE {?patient rdf:type McManus:HighRiskCholesterolRe...
Transparency!Reproducibility! Sharability! Comparability!Simplicity!Automation!Expert “tweaking” is allowed(the expert ret...
CAN WE INTERPRET COMPLEXCLINICAL PHENOTYPES?Problem #3: Moving Beyond Simple Binary Risk
The next step was to attempt to model theFramingham Risk Scorese.g. 10-year Cardiovascular Disease RiskThis takes a large ...
How do we do with these non-trivial cases?...not very well LOL!
OWL Modeling of Framingham 10-year risk for general CVDUGH! Awful!
Discussions with the clinical researcher revealed the problem...The patients were on drugs that affected their clinical me...
Can we compensate for that level of expert intuition?We believe soand the required knowledge is already encoded for us!
NDF-RT from the U.S. Veterans Authority
The resource to automate interpretationof a patient’s prescriptions existsand would allow us to (more) properlyinterpret t...
Patient1 Patient2DRUG 1 ASPRIN* ASCRIPTINDRUG 1 DOSAGE 1 DLY 1DLY,10MG AS NEEDEDDRUG 2 PROCARDIA PERSANTINEDRUG 2DOSAGE 10...
Patient1 Patient2DRUG 1 ASPRIN* ASCRIPTINDRUG 1 DOSAGE 1 DLY 1DLY,10MG AS NEEDEDDRUG 2 PROCARDIA PERSANTINEDRUG 2DOSAGE 10...
RxNav and ChemSpider have APIsfor canonicalization of drug namesUse SADI service workflow to migrate legacy datainto canon...
This workflow has a ~4% failure rate(my small trials with Google Suggest looked promisingat improving this!)
UnderHypertensionTreatment=galen:Patient andcardio:isPrescribed some(cardio:CanonicalizedDrugCollection andsio:has_member ...
Now how are we doing?The answer is a bit surprising...
Patient ID Automatic Risk Grade(based on drugsprescribed)Expert-assigned Grade(BP_TREATMENT_STATUS)Uri4627 1 1Uri4275 1 1U...
A large number of drugs are used to treat blood pressurebut these drugs can also used to treat other thingsFrom the perspe...
But the drug has other effectsthat the clinical researchermight not (does not) account forin their expert evaluationHow do...
Accuracy Precision RecallHigh Risk 0.82 0.89 0.71 0.84 0.30 0.61ModerateRisk0.68 0.73 0.68 0.72 0.71 0.74Low Risk 0.76 0.8...
We’re looking for other “intuitive” decisionsmade by the clinicianthat will account for our remaining inaccuraciesWe are o...
Interestingly, we were also able to create a simple OWL Classthat allowed classification of patients based on beingprescri...
Take-home messages - repeated1. We automated analysis using Semantic Web Services2. We encoded clinical guidelines in OWL3...
PublicationDiscourseHypothesisExperimentInterpretation??
The OWL Classes we constructedrepresent a particular clinician’s view of “reality”By my definition, that IS a hypothesis!O...
Life Web Science:The Web is a cradle-to-gravebiomedical research platform.
This is the work of Soroush SamadianPh.D. CandidateBioinformatics Programme, UBCVancouver, BC, Canada
Microsoft Research
Upcoming SlideShare
Loading in …5
×

Enhancing Reproducibility and Transparency in Clinical Research through Semantic Technologies

841 views

Published on

We were interested in whether we could model well-established clinical risk guidelines in OWL, and use these to automatically classify patient data v.v. "risk" (e.g. using the Framingham risk categories). What we ended-up doing, however, was wandering down a very interesting path of attempting to model clinical intuition! This reports the first phase of the experiment. A subsequent SlideShare will give part II of this investigation.

This is the work of Soroush Samadian, Ph.D. Candidate at the University of British Columbia Bioinformatics Graduate Programme.

Published in: Technology
  • Be the first to comment

Enhancing Reproducibility and Transparency in Clinical Research through Semantic Technologies

  1. 1. Life Web Science 2013, ParisImproving Transparency and Reproducibilityof Clinical ResearchUsing Semantic TechnologiesSoroush Samadian & Mark WilkinsonIsaac Peral Senior Researcher in Biological InformaticsCentro de Biotecnología y Genómica de Plantas, UPM, Madrid, SpainAdjunct Professor of Medical Genetics, University of British ColumbiaVancouver, BC, Canada.
  2. 2. Can we make the Web ascientific research platformfrom hypothesis right through to publication
  3. 3. Focus on publishing citable snippetsof scientific knowledge using SemWeb standards
  4. 4. That’s a good start v.v. academic publishing, but...What about the rest of the scientific process?
  5. 5. PublicationDiscourseHypothesisExperimentInterpretation
  6. 6. Houston, we have a problem...
  7. 7. ContextMultiple recent surveys of high-throughput biologyreveal that upwards of 50% of published studiesare not reproducible- Baggerly, 2009- Ioannidis, 2009
  8. 8. ContextSimilar (if not worse!) in clinical studies- Begley & Ellis, Nature, 2012- Booth, Forbes, 2012- Huang & Gottardo, Briefings in Bioinformatics, 2012
  9. 9. Context“the most common errors are simple,the most simple errors are common”At least partially because theanalytical methodology was inappropriateand/or not sufficiently described- Baggerly, 2009
  10. 10. ContextThese errors pass peer reviewThe researcher is unaware of the errorThe process that led to the error is not recordedTherefore it cannot be detected during peer-review
  11. 11. ContextDiscovery of such errors have resulted in retractionsand even shut-down clinical trials- Ioannidis, 2009
  12. 12. ContextIn March, 2012, the US Institute of Medicine ~said“Enough is enough!”
  13. 13. ContextInstitute of Medicine RecommendationsFor Conduct of High-Throughput Research:Evolution of Translational Omics Lessons Learned and the Path Forward. TheInstitute of Medicine of the National Academies, Report Brief, March 2012.1. Rigorously-described, -annotated, and -followed datamanagement and manipulation procedures2. “Lock down” the computational analysis pipeline once ithas been selected3. Publish the analytical workflow in a formal manner,together with the full starting and result datasets
  14. 14. How do we get there from here?
  15. 15. Our early attempts atsupporting clinical researchthrough SemWeb
  16. 16. MEASUREMENT UNITSProblem #1: Integrating Clinical Data
  17. 17. Units must be expressedUnits must be harmonizedDon’t leave this up to the researcher(it’s fiddly, time-consuming, and error-prone)
  18. 18. NASA Mars Climate Orbiter
  19. 19. Oops!
  20. 20. QUDT: Quantities, Units, Dimensions and Types Ontologyconversion offset & conversion multiplierenable conversion between non-SI-based unit and its SI equivalent.
  21. 21. OM: Ontology of Units of MeasurementUseful for inventing new units that are commonly used in clinical researchbut not (currently) in any Unit ontology
  22. 22. ID HEIGHT WEIGHT SBP CHOL HDL BMIGRSBPGRCHOLGRHDLGRpt1 1.82 177 128 227 55 0 0 1 0pt2 179 196 13.4 5.9 1.7 1 0 1 0Legacy clinical datasetused in our studiesHeight in m and cm Chol in mmol/l and mg/l...and other delicious weirdness Expert decision on “risk”(e.g. BMI=1 means “at health-risk with this BMI)
  23. 23. GOAL: get the clinical researcher“out of the loop” once the data is collectedComplete transparency in analysis withNO PEEKING & NO TWEAKING!(see U.S. IOM Recommendations)
  24. 24. Extending the GALEN ontology with richer logicincluding measurement values and unitsmeasure:SystolicBloodPressure =galen:SystolicBloodPressure and("sio:has measurement value" some "sio:measurement" and("sio:has unit" some “om: unit of measure”) and(“om:dimension” value “om:pressure or stress dimension”) and"sio:has value" some rdfs:Literal))Very general definition“some kind of pressure unit”
  25. 25. Now Galen classes can be used to “carry” rich dataMove beyond use of ontologies for simple keyword curation(keyword hierarchies are SO last-decade!)
  26. 26. Now, what do we do with the unit-soup that is in our legacy dataset?
  27. 27. SADI Semantic Web Service for automated Unit conversion• Send it a dataset with mixed units• (optional) tell it the harmonized unit you want back• Returns you a dataset with harmonized unitsAutomatic semantic detection of the “nature”of the incoming unit type (e.g. “unit of pressure”)Automatic conversion based on dimensionality and/or offset & multiplier
  28. 28. Create additional ontological classesrepresenting clinical features of interest based on clinical guidelinesmeasure:HighRiskSystolicBloodPressuremeasure:SystolicBloodPressure andsio:hasMeasurement some(sio:Measurement and(“sio:has unit” value om:kilopascal) and(sio:hasValue some double[>= "18.7"^^double])))Remember that this is fromour extension of GalenExtend, Reuse, Recycle!Now we’re being specificMUST be in kpascal and must be > 18.7
  29. 29. SELECT ?record ?convertedvalue ?convertedunitFROM <./patient.rdf>WHERE {?record rdf:type measure:HighSystolicBloodPressure .?record sio:hasMeasurement ?measurement.?measurement sio:hasValue ?convertedvalue.?record cardio:ExpertClassification ?riskgrade .}RecordID Start Val Start Unit End Val End Unitcm_hg1 15 cmHg 19.998 KiloPascalcm_hg2 14.6 cmHg 19.465 KiloPascalmm_hg1 14.8 mmHg 19.731 KiloPascalmm_hg2 146 mmHg 19.465 KiloPascalSHARE query(SHARE is a SADI-enhanced SPARQL query engine)Because the OWL definition of HighSBPrequired kpascal, SHARE used SADI toauto-convert everything into kpascal
  30. 30. ASSESSING RISKProblem #1: Automatic Interpretation of Clinical Data
  31. 31. Framingham risk measurements:AgeGenderHeightWeightBody Mass Index(BMI)Systolic Blood Pressure(SBP)Diastolic Blood Pressure(DBP)GlucoseCholesterolLow Density Lipoprotein (LDL)High Density Lipoprotein (HDL)Triglyceride(TG)All modeledas OWL Classesmuch the sameas described before
  32. 32. Measurements like BMI are derived from calculationsover other “core” measurementsAgain, we use SADI and semantics to achieve this automatically(and of course, any unit conflicts in the input data are also automaticallydetected and resolved by the previous SADI service we discussed)
  33. 33. Semantic Modeling of theAmerican Heart Association Risk gradesHighRiskBMI =PatientRecord and(sio:hasAttribute some(cardio:BodyMassIndex and sio:hasMeasurement some(sio:Measurement and(sio:hasUnit value cardio:kilogram-per-meter-squared) and(sio:hasValue some double[>= 25.0]))))Limit taken directly from clinical guidelinesSimilarly for SBP, Cholesterol, etc....
  34. 34. SHARE Query for High Risk SBP(SHARE is a SADI-enhanced SPARQL query engine)
  35. 35. Now the interesting question...
  36. 36. How does our automated risk evaluationcompare to a clinician’s expert risk classification?
  37. 37. True positive rate“at risk” %False positive rate“at risk”%SBP 100 0DBP 100 0CHOL 92.6 0HDL 100 56.5TG 100 8.5BMI 100 18.8LDL 100 0How does our automated risk evaluationcompare to a clinician’s expert risk classification?Yuck!
  38. 38. But... We encoded and were following the guidelines!We double-checked and our definitions were definitely correctHow could we possibly be wrong??
  39. 39. Visual inspection of data and guidelines showedin many cases the clinician had “tweaked” the guideline------------------AHA BMI risk threshold: BMI=25In our dataset the clinical researcher used BMI=26------------------HDL “official” guideline HDL=1.03mmol/lThe dataset from our researcher: HDL=0.89mmol/l-------------------
  40. 40. Adjusting our OWL class definitions and re-running the analysisResulted in nearly 100% correspondence with the clinical researcher(at least for binary risk assessment on simple measurements)HighRiskCholesterolRecord=PatientRecord and(sio:hasAttribute some(cardio:SerumCholesterolConcentration andsio:hasMeasurement some ( sio:Measurement and(sio:hasUnit value cardio:mili-mole-per-liter) and(sio:hasValue some double[>= 5.0]))))HighRiskCholesterolRecord=PatientRecord and(sio:hasAttribute some(cardio:SerumCholesterolConcentration andsio:hasMeasurement some ( sio:Measurement and(sio:hasUnit value cardio:mili-mole-per-liter) and(sio:hasValue some double[>= 5.2]))))
  41. 41. Reflect on this for a second... Because this is important!1. We automated data cleansing and analysis using Semantic Web Services2. We encoded clinical guidelines in OWL (first time this has been done AFAIK)3. We found that clinical researchers did not follow the official guidelines• This is fine! They’re the experts! But...4. Their “personalization” of the guidelines was unreported5. Nevertheless, we were able to create “personalized” OWL Classesrepresenting the viewpoint of that clinical researcher6. These personalized viewpoints, in OWL, were published on the Web7. These published, personalized OWL classes can be automatically re-usedby others to interpret their own data using that clinician’s viewpoint
  42. 42. AHA:HighRiskCholesterolRecordPatientRecord and(sio:hasAttribute some(cardio:SerumCholesterolConcentration andsio:hasMeasurement some ( sio:Measurement and(sio:hasUnit value cardio:mili-mole-per-liter) and(sio:hasValue some double[>= 5.0]))))McManus:HighRiskCholesterolRecordPatientRecord and(sio:hasAttribute some(cardio:SerumCholesterolConcentration andsio:hasMeasurement some ( sio:Measurement and(sio:hasUnit value cardio:mili-mole-per-liter) and(sio:hasValue some double[>= 5.2]))))PREFIX AHA=http://americanheart.org/measurements/PREFIX McManus=http://stpaulshospital.org/researchers/mcmanus/
  43. 43. To do the “experiment” using AHL guidelinesSELECT ?patient ?riskWHERE {?patient rdf:type AHL: HighRiskCholesterolRecord .?patient ex:hasCholesterolProfile ?risk}
  44. 44. To do the “experiment” using McManus guidelinesSELECT ?patient ?riskWHERE {?patient rdf:type McManus:HighRiskCholesterolRecord .?patient ex:hasCholesterolProfile ?risk}
  45. 45. Transparency!Reproducibility! Sharability! Comparability!Simplicity!Automation!Expert “tweaking” is allowed(the expert retains their expert authority)but these tweaks are explicit and transparent
  46. 46. CAN WE INTERPRET COMPLEXCLINICAL PHENOTYPES?Problem #3: Moving Beyond Simple Binary Risk
  47. 47. The next step was to attempt to model theFramingham Risk Scorese.g. 10-year Cardiovascular Disease RiskThis takes a large number of variables(SBP, BMI, and disease states such as diabetes)and calculates a patient’s risk
  48. 48. How do we do with these non-trivial cases?...not very well LOL!
  49. 49. OWL Modeling of Framingham 10-year risk for general CVDUGH! Awful!
  50. 50. Discussions with the clinical researcher revealed the problem...The patients were on drugs that affected their clinical measurements(effectively, the drugs made them more “normal”)however the expert continued to classify them as having the clinical problembased on their implicit knowledgeregardless of the clinical measurement value
  51. 51. Can we compensate for that level of expert intuition?We believe soand the required knowledge is already encoded for us!
  52. 52. NDF-RT from the U.S. Veterans Authority
  53. 53. The resource to automate interpretationof a patient’s prescriptions existsand would allow us to (more) properlyinterpret their phenotypeIFWe could accurately get this informationout of their clinical record
  54. 54. Patient1 Patient2DRUG 1 ASPRIN* ASCRIPTINDRUG 1 DOSAGE 1 DLY 1DLY,10MG AS NEEDEDDRUG 2 PROCARDIA PERSANTINEDRUG 2DOSAGE 10MG 1 3X DLY 75MG TIDDRUG 3 BUFFERIN LopidDRUG 3 DOSAGE 1DLY 4X300MG DLYDRUG 4 VASOTEC DICUMAROLDRUG 4 DOSAGE 2 DLYDRUG 5 XSD ASCRIPTIN TRANRENEDRUG 5 DOSAGEDRUG 6 DIPYRIDAMOLE100MGPERSANTINEDRUG 6 DOSAGE 1 75MG, 3X DLYDRUG 7 VASOTECDRUG 7 DOSAGETreated for HBP 1 1Treated for Diabetes 1 1Treated for HighCholesterol0 1
  55. 55. Patient1 Patient2DRUG 1 ASPRIN* ASCRIPTINDRUG 1 DOSAGE 1 DLY 1DLY,10MG AS NEEDEDDRUG 2 PROCARDIA PERSANTINEDRUG 2DOSAGE 10MG 1 3X DLY 75MG TIDDRUG 3 BUFFERIN LopidDRUG 3 DOSAGE 1DLY 4X300MG DLYDRUG 4 VASOTEC DICUMAROLDRUG 4 DOSAGE 2 DLYDRUG 5 XSD ASCRIPTIN TRANRENEDRUG 5 DOSAGEDRUG 6 DIPYRIDAMOLE100MGPERSANTINEDRUG 6 DOSAGE 1 75MG, 3X DLYDRUG 7 VASOTECDRUG 7 DOSAGETreated for HBP 1 1Treated for Diabetes 1 1Treated for HighCholesterol0 1
  56. 56. RxNav and ChemSpider have APIsfor canonicalization of drug namesUse SADI service workflow to migrate legacy datainto canonicalized form
  57. 57. This workflow has a ~4% failure rate(my small trials with Google Suggest looked promisingat improving this!)
  58. 58. UnderHypertensionTreatment=galen:Patient andcardio:isPrescribed some(cardio:CanonicalizedDrugCollection andsio:has_member some(cardio:HypertensionTreatmentMedication))cardio:HypertensionTreatmentMedication=cardio:CanonicalDrugRecord and ( ndf:may_treat some ndf:Hypertension )Adding prescription information into our OWLFramingham Risk modelsNote how easy it is to connect semantic data into your system -Just refer to it in your definition!! Also note that we’re not listing a bunch of drugs,we’re including any drug defined as a Hypertension treatment by NDF-RT.The Semantic Definition, not an explicit drug list!
  59. 59. Now how are we doing?The answer is a bit surprising...
  60. 60. Patient ID Automatic Risk Grade(based on drugsprescribed)Expert-assigned Grade(BP_TREATMENT_STATUS)Uri4627 1 1Uri4275 1 1Uri822 1 0Uri893 1 1For “treated for cholesterol” and “treated for diabetes”we achieved detection specificities of 96->99%But for blood pressure it was a bit more of a mess...Only 44% specificity! What went wrong?
  61. 61. A large number of drugs are used to treat blood pressurebut these drugs can also used to treat other thingsFrom the perspective of the treating clinicianthe purpose for which they prescribed the drugis the purpose that they record in the chart
  62. 62. But the drug has other effectsthat the clinical researchermight not (does not) account forin their expert evaluationHow do we define “correct” in this scenario?????(i.e. Is this a bug, or a feature??)
  63. 63. Accuracy Precision RecallHigh Risk 0.82 0.89 0.71 0.84 0.30 0.61ModerateRisk0.68 0.73 0.68 0.72 0.71 0.74Low Risk 0.76 0.83 0.55 0.65 0.80 0.81OverallOur ability to classify raw clinical data(with spelling mistakes and all)into the Framingham Risk evaluationcompared to the expert clinical assessmentWhite = before including prescription informationGrey = including the NDF-RT drug knowledgebase
  64. 64. We’re looking for other “intuitive” decisionsmade by the clinicianthat will account for our remaining inaccuraciesWe are optimistic that we can recordat least some of these in OWL and/oras features within the SPARQL queryRemember – the objective is transparencynot necessarily 100% semantic encoding
  65. 65. Interestingly, we were also able to create a simple OWL Classthat allowed classification of patients based on beingprescribed contra-indicated drugs~4.2% of patients were taking dangerous drug combinationsWe “got this for free” by connectinga bunch of semantic resources together!
  66. 66. Take-home messages - repeated1. We automated analysis using Semantic Web Services2. We encoded clinical guidelines in OWL3. We found that clinical researchers did not follow the officialguidelines• This is fine! They’re the experts! But...4. Their “personalization” of the guidelines was unreported5. We were able to create “personalized” OWL Classesrepresenting the viewpoint of that clinical researcher6. These personalized viewpoints were published on the Web7. The OWL classes can be automatically re-used by others
  67. 67. PublicationDiscourseHypothesisExperimentInterpretation??
  68. 68. The OWL Classes we constructedrepresent a particular clinician’s view of “reality”By my definition, that IS a hypothesis!Other work in our lab has demonstrated* that we canduplicate an entire published research paper“simply” by creating an OWL class representingthe hypothetical view of that researcher(and note that these hypotheses are explicit, shared on the Web,and re-usable by others!!)* Wood et al, Proc. ISoLA, 2012
  69. 69. Life Web Science:The Web is a cradle-to-gravebiomedical research platform.
  70. 70. This is the work of Soroush SamadianPh.D. CandidateBioinformatics Programme, UBCVancouver, BC, Canada
  71. 71. Microsoft Research

×