* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download Slide 1
Survey
Document related concepts
Transcript
Ontology Databases: Detecting Inconsistencies in the Gene Ontology using Not-gadgets Paea LePendu University of Oregon Talk: National Center for Biomedical Ontology ◦ Stanford University ◦ September, 2009 General Interests Logic Programming Languages Automated Reasoning Databases Outline • Ontology-based Data Management – Background, Motivation – Theory – Benchmarking – Application Domain, Query Answering • Inconsistency Detection – Theory – The serotonin example – GO plus ZFIN, MGI annotations Ontology-based Database Integration: reducing database integration to ontology translation Ontology-based Database Integration: reducing database integration to ontology translation Ontology-based Data Management Ontology-based Data Management User Ontology Data Access Layer Data Annotation Data Management RDBMS RDBMS RDBMS RDBMS RDBMS Example: sisters-siblings This is what we know : All sisters are siblings. Hilary and Lynn are sisters. This is what we want to know : Who are siblings? { <x,y> | siblingOf(x,y) } Obviously, the answer should be : Hilary and Lynn are siblings. { <Hilary, Lynn> } Example: sisters-siblings Example: sisters-siblings Example: The Gene Ontology GO_0003674 z01, z02, z03 GO_0005488 e01,e02, e03 GO_0030528 y01, y02, y03 GO_0003676 c01, c02, c03 GO_0045182 b01, b02, b03 GO_0003677 x01, x02, x03 GO_0003723 d01, d02, d03 GO_0003700 w01, w02, w03 GO_0008135 a01, a02, a03 Example: The Gene Ontology GO_0003674 z01, z02, z03 GO_0005488 e01,e02, e03 GO_0030528 y01, y02, y03 GO_0003676 c01, c02, c03 GO_0045182 b01, b02, b03 GO_0003677 x01, x02, x03 GO_0003723 d01, d02, d03 GO_0003700 w01, w02, w03 GO_0008135 a01, a02, a03 Ontology Databases: General Models for Database Designs • Generality is important – Avoid rewriting • Scalability of KB is important – Persistence, caching and indexing • Major generic models – Horizontal Models – Vertical Models – Decomposition Storage Models Ontology Databases: View-based Approach CREATE VIEW v_Person(id) AS SELECT id FROM Person UNION SELECT id FROM v_Male UNION SELECT id FROM v_Female v_Person Person P-0004 Person v_Female Female Male v_Male Female Male P-0002 P-0001 P-0003 [Pan & Heflin. DLDB: Extending Relational Databases to Support Semantic Web Queries. ISWC, 2003.] Ontology Databases: Active Database Approach ON INSERT into Male INSERT into Person On INSERT into Female INSERT into Person Person Person Female Male Female Male [LePendu, et al. Ontology Database: a New Method for Semantic Modeling and an Application to Brainwave Data. SSDBM, 2008.] Ontology Databases: Active Database Approach ON INSERT into Male INSERT into Person On INSERT into Female INSERT into Person Person Person Female Male Female Male P-0001 Ontology Databases: Active Database Approach Person P-0001 Person Female Male Female Male P-0001 Ontology Databases: Active Database Approach Person P-0001 Person Female Male Female Male P-0002 P-0001 Ontology Databases: Active Database Approach Person P-0001 P-0002 Person Female Male Female Male P-0002 P-0001 Ontology Databases: Active Database Approach Person P-0001 P-0002 Person Female Male Female Male P-0002 P-0001 P-0003 Ontology Databases: Active Database Approach Person P-0001 P-0002 P-0003 Person Female Male Female Male P-0002 P-0001 P-0003 Ontology Databases: Active Database Approach Person P-0001 P-0002 P-0003 Person P-0004 Female Male Female Male P-0002 P-0001 P-0003 Example: sisters-siblings (revisited) This is what we know : All sisters are siblings. Hilary and Lynn are sisters. This is what we want to know : Who are siblings? Obviously, the answer should be : Hilary and Lynn are siblings. Example: sisters-siblings (revisited) This is what we know : SiblingOf All sisters are siblings. Hilary and Lynn are sisters. This is what we want to know : Who are siblings? Obviously, the answer should be : Hilary and Lynn are siblings. SisterOf Hilary Lynn Example: sisters-siblings (revisited) This is what we know : SiblingOf Hilary Lynn All sisters are siblings. Hilary and Lynn are sisters. This is what we want to know : Who are siblings? Obviously, the answer should be : Hilary and Lynn are siblings. SisterOf Hilary Lynn Example: sisters-siblings (revisited) This is what we know : SiblingOf Hilary Lynn All sisters are siblings. Hilary and Lynn are sisters. SisterOf Hilary Lynn This is what we want to know : Who are siblings? Obviously, the answer should be : Hilary and Lynn are siblings. { <x,y> | siblingOf(x,y) } Example: sisters-siblings (revisited) This is what we know : SiblingOf Hilary Lynn All sisters are siblings. Hilary and Lynn are sisters. SisterOf Hilary Lynn This is what we want to know : Who are siblings? { <x,y> | siblingOf(x,y) } Just look it up! Obviously, the answer should be : Hilary and Lynn are siblings. Example: sisters-siblings (revisited) This is what we know : SiblingOf Hilary Lynn All sisters are siblings. Hilary and Lynn are sisters. SisterOf Hilary Lynn This is what we want to know : Who are siblings? { <x,y> | siblingOf(x,y) } Just look it up! Obviously, the answer should be : Hilary and Lynn are siblings. { <Hilary, Lynn> } Lehigh University Benchmark (LUBM) Load Time and Query Time (1.5 million facts) (10 Universities, 20 Departments) [Guo, et al. LUBM: A Benchmark for OWL Knowledge Base Systems. J Web Semantics, 2005.] Ontology-based Data Management [Frishkoff, et al. Development of Neural Electromagnetic Ontologies (NEMO): Ontology-based Tools for Representation and Integration of Event-related Brain Potentials. ICBO, 2009] Ontology-based Query Answering Return all data instances that belong to ERP pattern classes which have a surface positivity over frontal regions of interest and are earlier than the N400. Which patterns have a region of interest that is left-occipital and manifests between 220 and 300ms? What is the range of intensity mean for the region of interest for N100? Show the region of interest for all ERP patterns that occur between 0 and 300ms. Which PCA factor do P100 patterns most often appear in? What is the range of intensity mean for the region of interest for N100 patterns? Show the patterns whose region of interest is left occipital and occurs between 220 and 300ms. Inconsistency Detection • Background and Motivation – Expressiveness – From disjunctions to negations • Theory – Not-gadgets • Motivation – Serotonin example – ATP-gated cation channel activity • Results from ZFIN and MGI Annotations Not-gadgets ¬ → ¬ Example: inconsistency detection "Annotations in this way sometimes point to errors in the typetype relationships described in the ontology. An example is the recent removal of the type serotonin secretion as an is_a child of neurotransmitter secretion from the GO Biological Process ontology. This modification was made as a result of an annotation from a paper showing that serotonin can be secreted by cells of the immune system where it does not act as a neurotransmitter.“ [Hill, et al. Gene Ontology annotations: what they mean and where they come from. BMC Bioinformatics, 2008] Example: serotonin secretion gene-x not-gadget fail! gene-x Example: GO:0004931 ATP-gated cation channel activity (as of 3/09): [Term] id: GO:0004931 name: ATP-gated cation channel activity namespace: molecular_function def: "Catalysis of the transmembrane transfer of an ion by a channel that opens when extracellular ATP has been bound by the channel complex or one of its constituent parts." [GOC:mah, PMID:9755289] comment: Note that this term refers to an activity and not a gene product. Consider also annotating to the molecular function term 'purinergic nucleotide receptor activity ; GO:0001614'. synonym: "P2X activity" RELATED [] synonym: "purinoceptor" BROAD [] synonym: "purinoreceptor" BROAD [] is_a: GO:0005231 ! excitatory extracellular ligand-gated ion channel activity is_a: GO:0005261 ! cation channel activity Example: GO:0004931 GO:0004391 sub-graph (using Jambalaya): Example: GO:0004931 What is so interesting about GO:0004391? ZFIN ZDB-GENE-030319-2 p2rx2 NOT GO:0004931ZFIN:ZDBPUB-031031-8|PMID:14580944 IDA F purinergic receptor P2X, ligand-gated ion channel, 2 gene taxon:7955 20071005 ZFIN ZFIN ZDB-GENE-030319-2 p2rx2 GO:0004931ZFIN:ZDBPUB-031031-8|PMID:14580944 IGI ZFIN:ZDB-GENE-000427-3 F purinergic receptor P2X, ligand-gated ion channel, 2 gene taxon:7955 20071005 ZFIN Source: [1/13/2009] http://www.geneontology.org/gene-associations/ Example: GO:0004931 The not-gadget will raise a logical inconsistency. p2rx2 NOTGO:0004931 p2rx2 GO:0004931 GO_0004931 _GO_0004931 p2rx2 not-gadget fail! * Tables starting with an '_' are negations. Example: GO:0004931 GO:0004391 sub-graph (using Jambalaya): Example: GO:0004931 GO:0004391 sub-graph (using Jambalaya): Example: GO:0004931 GO:0004391 sub-graph (using Jambalaya): ZFIN ZFIN MGI MGI MGI ZFIN - MGI ZFIN Outcome: suspect IEA annotations GO Online SQL Environment (GOOSE) pos,IEA(graph_path x association) x neg(grapth_path x association) Source: [1/13/2009] http://www.geneontology.org/GO.database.shtml#diagram What do logical inconsistencies mean? • Several possibilities: – Incorrect annotation (e.g., suspect IEA annotations) – Incorrect relationship (e.g., serotonin secretion) – Incomplete model: Recall: ZFIN ZDB-GENE-030319-2 p2rx2 GO:0004931 ZFIN:ZDB-PUB-031031-8|PMID:14580944 IGI ZFIN:ZDB-GENE-000427-3 F purinergic receptor P2X, ligand-gated ion channel, 2 gene taxon:7955 20071005 ZFIN – Perfectly admissible! Next Directions • Explanation and proof-reconstruction ✓ • Deep (data) annotation tools • Distributed network of Ontology Databases Data Annotation: Neural ElectroMagnetic Ontologies frontocentral LFRON RFRON [Frishkoff, et al. ERP measures of partial semantic knowledge: Left temporal indices of skill differences and lexical quality. Biological Psychology, 2009.] Network of Ontology Databases [Thorisson, Muilu and Brookes. Genotype–phenotype databases: challenges and solutions for the post-genomic era. Nature Reviews, 2009.] Thank you Questions? Andrea’s Example Is John supervised by a TopManager who is a friend of an AreaManager? [Franconi. Ontologies and databases: myths and challenges. VLDB, 2008.] Raymond Reiter [Reiter. Deductive Question-Answering on Relational Data Bases. Logic and Data Bases, 1977] Raymond Reiter Raymond Reiter Raymond Reiter Benchmarking Suite Origins CIS @ UO CIS @ UO • Research Areas in Computer Science: – – – – – – – software engineering programming languages human-computer interaction parallel and distributed computing networking and graph theory scientific computation/visualization information integration and mining • Affiliates: – Neurosciences Institute – Computational Science Institute – Zebrafish Information Network Ontology-based Data Access [Rodriguez-Muro , et al. Realizing Ontology Based Data Access: A plug-in for protégé. ICDEW, 2008.]