Ci4SeR Curation Interface for Semantic Resources Evaluation with Adverse Drug Reactions SOUVIGNET Julien SOUVIGNET Julien, , ASFARI Hadyl, DECLERCK Gunnar, LARDON Jérémy, ASFARI Hadyl, DECLERCK Gunnar, LARDON Jérémy, TROMBERT-PAVIOT Béatrice, JAULENT Marie-Christine, BOUSQUET Cédric TROMBERT-PAVIOT Béatrice, JAULENT Marie-Christine, BOUSQUET Cédric INSERM 1142 – LIMICS, Paris, France CHU University of Saint-Etienne, France
Curatio Long-term Objective Interfac fo Semant Resource SOUVIGNE Julien - • Create and maintain an Ontology of Adverse Drug Reactions (OntoADR) INSERM U1142 – • Formally define MedDRA terms using SNOMED-CT LIMIC MIE 2014 – Istanbul – • Example: Septembe 201 “G G ASTRIC ASTRIC BLEEDING BLEEDING ” (M ED DRA) AXIOM 1 : HAS F INDING S ITE SOME “S S TOMACH TOMACH ” (SNOMED-CT) AXIOM 2 : HAS A SSOCIATED M ORPHOLOGY SOME “H H EMORRHAGE EMORRHAGE ” (SNOMED-CT) 2
Context Curation Interface for Se • Why MedDRA ? SOUVIGNET Julien - INSERM MIE 2014 – Istanbul – 1 • Standard for Adverse Event reporting • First generation terminology • 20k Preferred Terms • Why SNOMED-CT ? • Large medical coverage • 400k concepts • Previous studies have already tried to map MedDRA with SNOMED-CT 3
Curatio Research Status Interfac fo Semant Resource SOUVIGNE Julien - • One semantic resource: OntoADR INSERM U1142 – • 35k concepts LIMIC MIE 2014 – • 150k definitional axioms Istanbul – • Generated using multiple sources and through automatics methods Septembe 201 • UMLS • NLP Methods • Syntactic methods • 70% of MedDRA terms have at least one axiom 1) 1) Need Need for for other other mappings mappings and/or and/or definitions definitions 2) 2) Need Need for expert for expert review review and/or and/or curation curation 4
Objectives Curation Interface for Se • Manual review of MedDRA term definitions SOUVIGNET Julien - INSERM MIE 2014 – Istanbul – 1 • Limited to 800 concepts • Expert validation of the existing definition • Completion of partial definitions • Creation of definition for currently undefined terms • Previous attempts showed that: • Protégé is not suitable • Spreadsheets are impractical • Others softwares too annotation oriented Need Need for a for a dedicated dedicated Graphical Graphical User User Interface Interface 5
Curatio Software Specifications Interfac fo Semant Resource SOUVIGNE Julien - • Target users INSERM U1142 – • domain experts for the curation part LIMIC MIE 2014 – • any user for the consultation part Istanbul – Septembe 201 • Desired features • Search for a given term using id or string match; • Display of a list of concepts to define with optional user-defined list; • Display of a given term definition (list of axioms); • Validation/Invalidation of an axiom • Insertion of a new axiom, by selecting a property and a concept. • … 6
Ci4SeR 1 8 Curation Interface for Se 1. Search SOUVIGNET Julien - INSERM MIE 2014 – Istanbul – 1 2 2. List 3. Definition 3 4. Validation buttons 5. Axioms 4 6 insertion 6. Expert icon 7. Automatic 5 proposals 8. Comments 7 7
Curatio Expert work Interfac fo Semant Resource SOUVIGNE Julien - • Manual expert validation INSERM U1142 – • Medical dictionnaries & multiple sources LIMIC MIE 2014 – • 120 hours over 5 months Istanbul – • Work on first importance topics in Pharmacovigilance Septembe 201 • Bullous eruptions, acute renal failure, anaphylactic shock, rhabdomyolysis, aplastic anemia / pancytopenia, cardiac valve fibrosis, etc. • Expert role • Valid/Invalid • Improve (narrower or broader) • Enrich (new definitions) 8
Result example Curation Interface for Se SOUVIGNET Julien - INSERM “G G ASTRIC ASTRIC BLEEDING BLEEDING ” MIE 2014 – Istanbul – 1 HAS F INDING S ITE SOME “S S TOMACH TOMACH ” + HAS F INDING S ITE SOME “U U PPER PPER GASTROINTESTINAL GASTROINTESTINAL TRACT TRACT STRUCTURE STRUCTURE ” HAS A SSOCIATED M ORPHOLOGY SOME “H H EMORRHAGE EMORRHAGE ” + OCCURS A FTER SOME “G G ASTROINTESTINAL ASTROINTESTINAL ULCER ULCER ” + DUE T O SOME “H H ELICOBACTER ELICOBACTER PYLORI PYLORI GASTROINTESTINAL GASTROINTESTINAL TRACT TRACT INFECTION INFECTION ” + HAS D EFINITIONAL M ANIFESTATION SOME “A A BDOMINAL BDOMINAL PAIN PAIN ” + HAS D EFINITIONAL M ANIFESTATION SOME “M M ELENA ELENA ” + HAS D EFINITIONAL M ANIFESTATION SOME “V V OMIT OMIT : : BLOOD BLOOD PRESENT PRESENT ” 9
Curatio Preliminary Results Interfac fo Semant Resource SOUVIGNE Julien - • 783 terms defined INSERM U1142 – Before After • 1369 properties validated LIMIC MIE 2014 – • 488 invalidated properties Istanbul – Approx. 2.63 relations Approx. 8.45 relations • 5333 new properties Septembe per definition in average per definition in average 201 • 80 narrower & 12 broader 10
Preliminary Results Curation Interface for Se • Before Ci4SeR, our team SOUVIGNET Julien - INSERM MIE 2014 – Istanbul – 1 already tried to define MedDRA terms manually Before After +486% +68% 11
Curatio Discussion & Perspectives Interfac fo Semant Resource SOUVIGNE Julien - • OntoADR is based on partial definitions INSERM U1142 – • Lots of new axioms LIMIC MIE 2014 – • Some invalid definitions Istanbul – Septembe 201 • Interface is time-saving • Time spent : 783 terms defined in 5 month (not full time) • Interface encourages completeness and accuracy • Need for collaboration with other experts 12
Thank you for your attention ! Ci4SeR Curation Interface for Semantic Resources Evaluation with Adverse Drug Reactions SOUVIGNET Julien SOUVIGNET Julien, , ASFARI Hadyl, DECLERCK Gunnar, LARDON Jérémy, ASFARI Hadyl, DECLERCK Gunnar, LARDON Jérémy, TROMBERT-PAVIOT Béatrice, JAULENT Marie-Christine, BOUSQUET Cédric TROMBERT-PAVIOT Béatrice, JAULENT Marie-Christine, BOUSQUET Cédric INSERM 1142 – LIMICS, Paris, France CHU University of Saint-Etienne, France
Recommend
More recommend