1. Ontological Model for Colon Carcinoma: A Case Study for Knowledge Representation in Clinical Bioinformatics Kumar A 1,2, Yip L 3, Jaremek M 2, Scheib H 3 1 IFOMIS, University of Saarland, Germany 2 Ludwig-Maximilians Universität, München, Germany 3 Swiss Institute for Bioinformatics, Geneva, Switzerland ONCOLOGYONTOLOGYONCOLOGYONTOLOGY

2. How the two worlds meet? ONCOLOGYONTOLOGYONCOLOGYONTOLOGY Clinical Molecular Biology Specific disease topics E-Health, Health support system Patient Management Patient health education Biomedical Information on the Web Swiss-Prot, Swiss-Prot Variant pages Proteins, mutations, functions and structures ?

3. How the two worlds meet? ONCOLOGYONTOLOGYONCOLOGYONTOLOGY Diseases  Pathological Processes  Body site for diseases  Diseases by staging  Risk factors Anatomy  Is-a, part-of  Granular relationship Biological Processes  Ontology  Classification Swiss-Prot proteins  Annotation: function, structure, mutation SNOMED GO FMA

4. How the two worlds meet? ONCOLOGYONTOLOGYONCOLOGYONTOLOGY ModSNP database Protein 3D models Swiss-Prot entries Disease schema

5. What is Protégé? Frame based system : Allows formation of class- subclass relations, provides support for other relations between classes, compatible with OWL, XML, database standards Support for various types of visualizations : Graphical, Web-based Support for import/export Support for reasoning : Description Logic based Can give outputs in various formats Difficulties with input ONCOLOGYONTOLOGYONCOLOGYONTOLOGY

6. Disease Representation Disease classification based on Snomed CT Various aspects considered for classification (currently present in the form of multiple inheritance) Added from textbooks (deVita Principles of Oncology and Harrison Principles of Internal medicine) –Staging of diseases (TNM, Duke’s, Modified Asler-Coller) –Screening (Patients screened based on their level of risk) –Risk factors –Localization –Pathology –To be added: Pharmacotherapeutics, Symptoms and Signs ONCOLOGYONTOLOGYONCOLOGYONTOLOGY

7. Anatomical and Histological Representation Anatomy of colon represented at Organ system, Organ, Tissue, Cell and Subcellular levels of granularity (Foundational Model of Anatomy) Gene Ontology‘s Cellular Component axis situated within the FMA axis Gross pathology mapped to the Carcinoma location Information regarding Clinical procedures, Carcinoma extent, Vascular invasion, Histological pathology being added Extensions being done to add relations like is-located- in, is-surrounded-by, etc. to make the anatomical representation deducible ONCOLOGYONTOLOGYONCOLOGYONTOLOGY

8. Interlink between disease, LocusLink, Swiss- Prot, GO annotations ONCOLOGYONTOLOGYONCOLOGYONTOLOGY Colon cancer/Colon carcinoma LocusLink SwissProtGene Ontology (246 distinct loci)

9. Gene Ontology Association rules found considering the Gene Ontology annotations of SWISSPROT proteins Gene Ontology consists of three axes: –Cellular Component –Molecular Function –Biological Processes Association between GO terms were established on the basis of these annotations –Database-based approach –Apriori-algorithm based approach –Dependency relations based on POS tagger ONCOLOGYONTOLOGYONCOLOGYONTOLOGY

10. Levels of granularity Levels of granularity in human body –Organism –Organ system –Cardinal body parts –Organ –Organ part –Tissues –Cells –Subcellular organelles –Molecules –Atoms Fundamentals behind the levels of granularity –Grains, Structure, Origin ONCOLOGYONTOLOGYONCOLOGYONTOLOGY

11. Results ONCOLOGYONTOLOGYONCOLOGYONTOLOGY RNA bindingnucleus30 DNA bindingnucleosome31 transcriptional activator activitynucleus31 structural constituent of ribosomecytosolic large ribosomal subunit (sensu Eukarya)32 transmembrane receptor activityintegral to plasma membrane36 protein bindingcytoplasm36 zinc ion bindingnucleus43 protein bindingnucleus45 receptor activityintegral to plasma membrane56 G-protein coupled receptor activityintegral to plasma membrane70 DNA bindingnucleus100 antigen bindingextracellular123 transcription factor activitynucleus171

12. Results (Apriori) Support and Confidence are defined ribosome <- ribosome biogenesis; protein biosynthesis (0.2%, 93.2%) This rule says that there are 0.2% of the total annotations, put together ribosome biogenesis and protein biosynthesis, of which 93.2% (i.e. 82) are also annotated with the term ribosome. Formal ontological relationsapplied between the entities, which would help to have deductions: has spatial projection, processual part of, facilitates, mediates, perpetrates ONCOLOGYONTOLOGYONCOLOGYONTOLOGY

13. Relations to biological pathways Pathway resources: KEGG, PATH, GenePath Protein interaction database resources: DIP, BIND, PiP, IntAct Links to these databases possible through Swissprot and GO annotations Associations found within GO annotations parallel the pathways and protein interactions (under verification) Text mining resources being considered ONCOLOGYONTOLOGYONCOLOGYONTOLOGY

14. Ontological Model for Colon Carcinoma: A Case Study for Knowledge Representation in Clinical Bioinformatics Kumar A 1,2, Yip L 3, Jaremek M 2, Scheib H 3 1 IFOMIS, University of Saarland, Germany 2 Ludwig-Maximilians Universität, München, Germany 3 Swiss Institute for Bioinformatics, Geneva, Switzerland ONCOLOGYONTOLOGYONCOLOGYONTOLOGY