R T U New York State Center of

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Clinical Trial Ontology Meeting How to build an Ontology ? Some basic principles NIH, May 16 -17, 2007 Werner CEUSTERS, MD Center of Excellence in Bioinformatics and Life Sciences Department of Psychiatry, University at Buffalo, NY, USA http: //www. org. buffalo. edu/RTU

R T U New York State Center of Excellence in Bioinformatics & Life Sciences 1977 1959 - 2006 Short personal history 1989 2004 1992 2002 1998 2

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Mainstream interpretations of “ontology” • An explicit specification of an agreed upon conceptualization of a domain – Tom Grüber • Anything what is given the name ‘ontology’ and that can be described in terms of 6 axes: expressiveness, structure, intended use, granularity, automated reasoning, prescriptive/descriptive – Ontology Summit 2007 3

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Problems with mainstream ontologies • Based upon the confusing notion of “concept” – Unit of thought or knowledge concerning anything perceivable or conceivable – The meaning of a term –… • Confuse information representation with domain representation Information about X part_of information about Y X part of Y 4

R T U New York State Center of Excellence in Bioinformatics & Life Sciences What I mean with the word “ontology” • A representation of some pre-existing domain of reality (a portion of reality) which 1. reflects the properties of the entities within its domain in such a way that there obtains a systematic correlation between reality and the representation itself, 2. is intelligible to a domain expert 3. is formalized in a way that allows it to support automatic information processing 5

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Three levels of reality 1. The world exists ‘as it is’ prior to a cognitive agent’s perception thereof; 2. Cognitive agents build up ‘in their minds’ cognitive representations of the world; 3. To make these representations publicly accessible in some enduring fashion, they create representational artifacts that are fixed in some medium. Smith B, Kusnierczyk W, Schober D, Ceusters W. Towards a Reference Terminology for Ontology Research and Development in the Biomedical Domain. Proceedings of KR-MED 2006, November 8, 2006, Baltimore MD, USA 6

R T U New York State Center of Excellence in and is relevant Represent what exist Bioinformatics & Life Sciences RU 1 O 1 B Cognitive representation concretization O R 1 st level reality 7

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Some characteristics of representational units 1. each unit is assumed by the creators of the representation to be veridical, i. e. to conform to some relevant POR as conceived on the best current scientific understanding; 2. several units may correspond to the same POR by presenting different though still veridical views or perspectives; 3. what is to be represented by the units in a representation depends on the purposes which the representation is designed to serve. 8

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Some characteristics of an optimal ontology • Each representational unit in such an ontology would designate – (1) a single portion of reality (POR), which is – (2) relevant to the purposes of the ontology and such that – (3) the authors of the ontology intended to use this unit to designate this POR, and – (4) there would be no PORs objectively relevant to these purposes that are not referred to in the ontology. 9

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Three types of ontologies • Upper level ontologies: – (should) describe the most generic structure of reality • Domain ontologies: – (should) describe the portion of reality that is dealt with in some domain – Special case: reference ontologies • Application ontologies: – To be used in a specific context and to support some specific application 10

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Clinical trial ontologies • As domain ontologies: – Cover all entity types relevant in the clinical trial domain • As application ontologies: – A subset of the above which is large enough to support all functions the application has to serve: • • CT protocol development Study management Data analysis … 11

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Key question How to build an optimal clinical trial domain ontology ?

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Rule 1: Analyze the domain

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Rule 2 a: Try to be lazy: re-use what others have done.

R T U New York State Center of Excellence in Bioinformatics & Life Sciences The BRIDG (domain analysis) model • NOT an ontology • A computable clinical trials protocol representation – that supports the entire life-cycle of clinical trial protocols, and – that will serve as a foundation for ca. BIG modules • that support all phases of the clinical trials life cycle, (including protocol authoring) and • be developed to meet user needs and requirements. The BRIDG Project: Creating a model of the semantics of clinical trials research. Douglas B. Fridsma. July 26, 2006 15

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Reasons for selecting BRIDG • BRIDG tries to solve an important problem • Does not completely ignore reality as many other initiatives do: (although one has to search hard to find evidence and sometimes it looks as if some contributors observed reality from outer space) – If the tools and models don’t work with reality, it is probably the tools and the models that need to change • The BRIDG Project: Creating a model of the semantics of clinical trials research. Douglas B. Fridsma. July 26, 2006 • Intended to become the next best thing on earth (after HL 7, I assume) 16

R T U New York State Center of Excellence in Bioinformatics & Life Sciences http: //www. bridgproject. org/status. html 17

R T U New York State Center of Excellence in Bioinformatics & Life Sciences BRIDG_Model_V 1_49 18

R T U New York State Center of Excellence in Bioinformatics & Life Sciences BRIDG model organization Image from: The BRIDG Project: Creating a model of the semantics of clinical trials research. Douglas B. Fridsma. July 26, 2006 19

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Rule 2 b: Try to be lazy: re-use what others have done, But… remain critical at all times!

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Being critical ≠ being negative RFQ-NCI-60001 -NG: Review of NCI Thesaurus and Development of Plan to Achieve OBO-Compliance Grant to Apelon (H. Solbrig) to improve NCIT 21

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Rule 3: Don’t have a blind trust in the power of representation and modeling languages, and certainly not in UML

R T U New York State Center of Excellence in Bioinformatics & Life Sciences ‘Death by UML Fever’ • It is important to emphasize that UML itself is not the direct cause of any maladies described herein. • Instead, UML is largely an innocent victim caught in the midst of poor process, no process, or sheer incompetence of its users. • UML sometimes does amplify the symptoms of some fevers as the result of the often divine-like aura attached to it. • For example, it is not uncommon for people to believe that no matter what task they may be engaged in, mere usage of UML somehow legitimizes their efforts or guarantees the value of the artifacts produced. Alex E. Bell. Death by UML Fever. Queue 2(1), March 2004, ACM Press, 72 – 80, 2004 23

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Who would not be impressed ? • Fig. 10: BRIDG Comprehensive Class and attribute diagram - (Logical diagram), p 99 24

R T U New York State Center of Excellence in Bioinformatics & Life Sciences I’m not ! • I have come to appreciate domain modeling in UML as an implementation-independent approach which is more likely to uncover “the truth” about the underlying semantics. – Dr. Diane Wold. Modeling Trial Design with BRIDG. July 26, 2006 • The UML diagram helped us to keep separate an activity, which exists independent of any schedule, and an activity-at-a-visit, (the X), which is a plan to perform that activity at a particular time. 25

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Rule 4: Limit the number of developers/contributors

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Contributors to the BRIDG model A chain is as strong as its weakest link Image from: The BRIDG Project: Creating a model of the semantics of clinical trials research. Douglas B. Fridsma. July 26, 2006 27

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Rule 5: • Be consistent in what you describe: – either representational units, or – the entities represented by them. • Thus: keep the levels of reality all the time in mind

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Living. Subject (BRIDG logical model p 1031) • • Type: Class Status: . Version. Phase. Package: Entities and Roles Keywords: Detail: Created on 02/09/2006. Last modified on 02/09/2006. • GUID: {7 C 04 F 8 D 8 -30 B 9 -4942 -B 2 A 8 -4 CF 93 E 8913 D 9} • An object representing an organism or complex animal, alive or not. Examples: person, dog, microorganism, plant of any taxonomic group, tissue sample, bacteria, fungi, and viruses. 29

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Substance. Administration (BRIDG logical model p 84) • • Type: Class Performed. Activity Status: Proposed. Version 1. 0. Phase 1. 0. Package: CTOM Elements Keywords: Detail: Created on 01/05/2005. Last modified on 12/14/2006. • GUID: {2289 C 0 E 8 -855 D-42 e 3 -86 FA 2 ECBE 59 D 8982} • The description of applying, dispensing or giving agents or medications to subjects. 30

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Person (BRIDG logical model p 106 a. f. , HE!) • • Type: Class Status: Proposed. Version 1. 0. Phase 1. 0. Package: Clinical Research Entities Keywords: Detail: Created on 06/09/2005. Last modified on 01/13/2007. • GUID: {6 F 49 F 110 -7 B 36 -4 c 03 -A 7 EAF 456 CE 1 E 739 D} • A human being. 31

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Some Person Attributes • administrative. Gender. Code (p 107) – The classification of the sex or gender role of the patient. Values include: Female, Male, and Unknown. • gender. Code (p 108) – The text that describes the assemblage of physical properties or qualities by which male is distinguished from female; the physical difference between male and female within a person. [Explanatory Comment: Identification of sex is usually based upon self-report and may come from a form, questionnaire, interview, etc. ] 32

R T U New York State A better example: Clinical Trial Ontology under DOLCE Center of Excellence in Bioinformatics & Life Sciences Crenguta Bogdan, Daniela Luzi, Fabrizio L. Ricci, Luca D. Serbanati. Towards a Clinical Trial Ontology using a Concern-Oriented Approach. W. P. n. 10, October 2006. 33

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Rule 6: Use a Realism-based Upper Ontology to classify the representational units in your Domain Ontology

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Realism in Basic Formal Ontology (BFO) • The world consists of – entities that are • Either particulars or universals; • Either occurrents or continuants; and, – relationships between these entities of the form • Either dependent or independent; • <particular , universal> • <particular , particular> • <universal , universal> e. g. is-instance-of, e. g. is-member-of e. g. isa (is-subtype-of) Smith B, Kusnierczyk W, Schober D, Ceusters W. Towards a Reference Terminology for Ontology Research and Development in the Biomedical Domain. Proceedings of KR-MED 2006, November 8, 2006, Baltimore MD, USA 35

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Only what exists (or existed) can be represented • Anything else can be imagined • Examples of what exist: – – – Body parts Disorders Abortions Women with prevented abortions Plans about my future activities • What does not exist – Prevented abortions – My future activities 36

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Planned. Activity (BRIDG logical model p 202, HE!) 37

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Rule 7: • Use formal ontological methods to: – – distinguish distinct entities assess in what way distinct entities are distinct

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Organism (BRIDG logical model p 160, HE!) • • Type: Class Status: Proposed. Version 1. 0. Phase 1. 0. Package: Clinical Research Roles Keywords: Detail: Created on 12/13/2006. Last modified on 01/19/2007. • GUID: {B 9 F 321 DB-365 F-4155 -B 8 F 6 -3 D…. • The role that a biological entity has, and that role participates in a microbiology test in two ways: first, it can be identified as the result of a microbiology test. It can also participate as a specimen in the microbiology test. [HL 7 Perspective] 39

R T U New York State Center of Excellence in Bioinformatics & Life Sciences An example: ONTOCLEAN • Identity, essence, unity, dependence C. Welty, N. Guarino"Supporting ontological analysis of taxonomic relationships", Data and Knowledge Engineering vol. 39, no. 1, pp. 51 -74, 2001 40

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Rule 8: • Don’t confuse reality with our means to access that reality, f. i. : • Don’t confuse the observation of an entity with the entity observed

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Adverse. Event (BRIDG logical model p 168, HE!) • • Type: Class Assessment Status: Proposed. Version 1. 0. Phase 1. 0. Package: Clinical Research Activities Keywords: Detail: Created on 05/24/2006. Last modified on 01/26/2007. • GUID: {CD 620136 -3 CB 9 -4382 -802 B-F 6 CA 82 F 98 C 10} • An observation of a change in the state of a subject that is assessed as being untoward by one or more interested parties within the context of protocol-driven research or public health. 42

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Example: medical ‘findings’ and ‘observations’ • A particular pathological entity may at a certain time be undetectable by any observation method or technique available to an observer, including the person exhibiting the pathological entity itself. 43

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Example: medical ‘findings’ and ‘observations’ (1) • • A particular pathological entity may at a certain time be undetectable by any observation method or technique available to an observer, including the person exhibiting the pathological entity itself. A particular observation (‘act of looking’) may produce false results and thus simulate the existence of a pathological entity. 44

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Example: medical ‘findings’ and ‘observations’ (1) • • • A particular pathological entity may at a certain time be undetectable by any observation method or technique available to an observer, including the person exhibiting the pathological entity itself. A particular observation may produce false results and thus simulate the existence of a pathological entity. An observer may observe or fail to observe a detectable particular pathological entity. 45

R T U New York State Center of Excellence in Bioinformatics & Life Sciences On ‘findings’ and ‘observations’ (2) • When an observer perceives a particular pathological entity, he might judge it – (1) to be an instance of the universal of which it is indeed an instance in reality, – (2) to be an instance of another universal (and thus be in error), or – (3) he might be not able to make an association with any universal at all. • Distinct manifestations of ‘the same type’ may be pathological or not: – Singing naked under the shower versus in front of The White House • . . . 46

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Rule 9: Do not accept silly suggestions, whomever they come from

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Device (BRIDG logical model, p 100, HE!) • • Type: Class Material Status: Proposed. Version 1. 0. Phase 1. 0. Package: Clinical Research Entities Keywords: Detail: Created on 02/22/2006. Last modified on 01/04/2007. • GUID: {3546 A 977 -C 51 F-4860 -A 09 A-2 ADAE 896 D 74 B} • <PROPOSED> A therapeutic or diagnostic intervention utilizing a piece of equipment or a mechanism designed to serve a special purpose or perform a special function whose basic characteristics are not altered in the course of the intervention. 48

R T U New York State Center of Excellence in Bioinformatics & Life Sciences The latter could also go under other rules: • • Stop working when you are tired Be careful with cut and paste Proof-read your work … 49

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Rule 10: Use distinct names for distinct representational units that denote distinct entities

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Adverse. Event (BRIDG logical model p 504) • • Type: Class Health. Problem Status: Proposed. Version 1. 0. Phase 1. 0. Package: Adverse Event Keywords: Detail: Created on 05/01/2006. Last modified on 05/02/2006. • GUID: {6783 F 6 F 2 -8837 -4 b 7 d-B 81 BA 25206 D 36689} • A toxic reaction to a medical therapy, or to an experience such as consuming a meal. 51

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Adverse. Event (BRIDG logical model p 91) • • • Type: Class Status: Proposed. Version 1. 0. Phase 1. 0. Package: SDTM Keywords: Detail: Created on 12/14/2005. Last modified on 12/28/2006. GUID: {F 1786 F 01 -F 973 -426 d-B 765 -0107 B 5823 A 18} Any untoward medical occurrence in a patient or clinical investigation subject administered a pharmaceutical product and which does not necessarily have a causal relationship with this treatment. An adverse event (AE) can therefore be any unintended sign (including an abnormal laboratory finding), symptom, or disease temporally associated with the use of a medicinal (investigational) product, whether or not related to the medicinal investigational) product. 52

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Adverse. Event (BRIDG logical model p 36) • • • Type: Class Assessment Status: Proposed. Version 1. 0. Phase 1. 0. Package: CTOM (imported package) Keywords: Detail: Created on 01/05/2005. Last modified on 09/26/2005. GUID: {C 0 F 30 FE 6 -EE 1 E-443 e-A 7 AB-256342 B 193 B 3} An unfavorable and unintended reaction, symptom, syndrome, or disease encountered by a subject while on a clinical trial regardless of whether or not it is considered related to the product or procedure. . The concept refers to assessments that could be medically related, dose related, route related, patient related, caused by an interaction with anotherapy or procedure, or dose escalation. 53

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Rule 11: Avoid contradictions

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Objective. Result (BRIDG logical model p 191, HE!) • • Type: Class Investigative. Result, Observation Status: Proposed. Version 1. 0. Phase 1. 0. Package: Clinical Research Activities Keywords: Detail: Created on 01/20/2005. Last modified on 12/28/2006. • GUID: {F 388 CFB 0 -77 DE-4008 -B 222 -EB… • An act of monitoring, recognizing and noting reproducible measurement of some magnitude with suitable instruments or established scientific processes. • <EXAMPLE> A laboratory test with standardized instruments, ECG measurement or question on a validated questionnaire such as SF 36. 55

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Some attributes of Objective. Result • missed. Indicator boolean – This is an indicator flag that flags a performed observation as "not done". (default: CDISC) …… p 193 • missed. Reason – This captures SDTM's ---REASND. In HL 7, there is a list of permissible missing value types, and we need to ensure that HL 7's list is a superset of what is needed by SDTM. – <EXAMPLE> A planned observation was not done because the equipment failed, so the corresponding "performed observation" exists as a placeholder to describe why that performed observation was not done……. ……p 193 – Default: [CDISC SDTM IG v 3. 1. 1 = REASND ] 56

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Rule 12: Avoid circular definitions

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Ingredient (BRIDG logical model p 507) • Status: Proposed. Version 1. 0. Phase 1. 0. • Package: Adverse Event Keywords: • Detail: Created on 03/01/2006. Last modified on 03/01/2006. • GUID: {7 D 53 B 2 A 1 -CEC 4 -49 ae-8 BD 6611 E 2 CF 4 D 862} • A substance that acts as an ingredient within a product. Note, that ingredients may also have ingredients. 58

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Rule 13: Do not use names with a precise meaning in general language to designate entities which are of a more specific or totally different type in the context of a specific application

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Animal (BRIDG logical model p 526) • • Type: Class Investigated. Party Status: Proposed. Version 1. 0. Phase 1. 0. Package: Investigated. Subject Keywords: Detail: Created on 03/10/2006. Last modified on 03/10/2006. • GUID: {996 CB 91 C-04 EC-4 b 1 d-9 AFF-57 B 878 D 532 D 7} • A non-person living entity which is chosen to be the subject of an investigation, or which is the subject of an • implicated act. 60

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Rule 14: Provide a mechanism to let the ontology evolve in line with changes in reality and in our understanding thereof

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Reality versus beliefs, both in evolution t U 1 U 2 Reality p 3 IUI-#3 Belief O-#0 O-#2 O-#1 = “denotes” = what constitutes the meaning of representational units …. Therefore: O-#0 is meaningless 62

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Changes in reality, beliefs, representations t U 1 U 2 R p 3 IUI-#3 B O-#0 O-#2 O-#1 Relationships amongst universals (R) or beliefs therein (B) 63

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Mistakes, discoveries, being lucky, having bad luck Mistakes t U 1 U 2 R p 3 IUI-#3 B O-#0 O-#2 O-#1 64

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Mistakes, discoveries, being lucky, having bad luck t U 1 U 2 R p 3 IUI-#3 B O-#0 O-#2 O-#1 65

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Mistakes, discoveries, being lucky, having bad luck t U 1 U 2 R p 3 IUI-#3 B O-#0 O-#2 O-#1 66

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Mistakes, discoveries, being lucky, having bad luck t U 1 U 2 R p 3 IUI-#3 B O-#0 O-#2 O-#1 67

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Key requirement for versioning Any change in an ontology or data repository should be associated with the reason for that change to be able to assess later what kind of mistake has been made ! 68

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Example: a person’s gender in the EHR • In John Smith’s EHR: – At t 1: “male” at t 2: “female” 69

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Example: a person’s gender in the EHR • In John Smith’s EHR: – At t 1: “male” at t 2: “female” • What are the possibilities ? 70

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Example: a person’s gender in the EHR • In John Smith’s EHR: – At t 1: “male” at t 2: “female” • What are the possibilities ? • Change in reality: 71

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Example: a person’s gender in the EHR • In John Smith’s EHR: – At t 1: “male” at t 2: “female” • What are the possibilities ? • Change in reality: • transgender surgery 72

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Example: a person’s gender in the EHR • In John Smith’s EHR: – At t 1: “male” at t 2: “female” • What are the possibilities ? • Change in reality: • transgender surgery • change in legal self-identification 73

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Example: a person’s gender in the EHR • In John Smith’s EHR: – At t 1: “male” at t 2: “female” • What are the possibilities ? • Change in reality: • transgender surgery • change in legal self-identification • Change in understanding: it was female from the very beginning but interpreted wrongly 74

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Example: a person’s gender in the EHR • In John Smith’s EHR: – At t 1: “male” at t 2: “female” • What are the possibilities ? • Change in reality: • transgender surgery • change in legal self-identification • Change in understanding: it was female from the very beginning but interpreted wrongly • Correction of data entry mistake: it was understood as male, but wrongly transcribed 75

R T U New York State Center of Excellence in Bioinformatics & Life Sciences Conclusion • Building high quality ontologies is hard. • Experts in driving cars are not necessarily experts in car mechanics (and the other way round). – Good computer scientists are usually lousy ontologists • Ontologies should represent the state of the art in a domain, i. e. the science. – Science is not a matter of consensus or democracy. • Natural language relates more to how humans talk about reality or perceive it, than to how reality is structured. • No high quality ontology without the involvement of ontologists. 76