Semantics for Bioinformatics: What, Why and How of Search, Integration and Analysis


Published on

Amit Sheth's Keynote at Semantic Web Technologies for Science and Engineering Workshop (held in conjunction with ISWC2003), Sanibel Island, FL, October 20, 2003.

Published in: Education, Technology
  • Be the first to comment

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide
  • Semantics (of information, communication) is a very old area, and extensive work on Semantic Technology has been going on for well over a decade (many projects on semantic interoperability, semantic information brokering) Semantic Web and related visions are being achieved in various depth and scope – mostly starting with targeted applications where requirements are much better understood and scope is manageable
  • GO (Gene ontology). KEGG (Kyoto Encyclopedia of Genes and Genomes) is a bioinformatics resource for understanding higher order functional meanings and utilities of the cell or the organism from its genome information. TAMBIS (Transparent Access to Multiple Bioinformatics Information Source). TAMBIS aims to aid researchers in biological science by providing a single access point for biological information sources round the world. EcoCyc , a part of the BioCyc library, is a scientific database for the bacterium Escherichia coli. The EcoCyc project performs literature-based curation of the entire E. coli genome, and of E. coli transcriptional regulation, transporters, and metabolic pathways. BioPAX (Biological Pathways Exchange).
  • Go ontology (schema) – corresponding KB for gene interaction can come from: Protein-protein interaction, GenBank phylogenetic relatedness, micro array data coexpression
  • Is gene the biologist researching in other organism? Is a similar gene in other organism? [if annotated, directly access DB, if not, use BLAST to normalize] Finding potential antifungal drug targets (most known antifungals are associated with Sterol metabolism), using BLAST to normalize the results (genetic sequences) from different databases associated with different organisms SGD: baker’s yeast; GUS: human, plasmodium, trypanosomes, ..; FGDB: pneumocystis Services: BLAST, co expression analysis, phylogeny
  • Change all these things.. Make them more concept killing
  • Big Question: What essential genes we all have in common?
  • Big Question: What essential genes we all have in common?
  • Essential eukaryotic core. All 202 genes of P. carinii listed are shared between P. carinii , S. pombe , and S. cerevisiae . All genes included are either lethal as knockouts in S. pombe or S. cerevisiae . Most genes on the diagram use the S. cerevisiae name, but a few follow the naming convention of S. pombe ( i.e ., ypt2 ), when there was more annotation in S. pombe . Genes in DNA- (light blue), RNA- (dark blue), protein- (red), signaling- (purple), metabolism- (orange), transport- (green), or other- (black) related processes are color-coded.
  • Semantics for Bioinformatics: What, Why and How of Search, Integration and Analysis

    1. 2. Acknowledgements <ul><li>UGA Biologists/Biochemists </li></ul><ul><li>Will York , Complex Carbohydrate Research Center </li></ul><ul><li>Jonathan Arnold , Fungal Genomics </li></ul><ul><li>Philip Bowen , CCQC </li></ul><ul><li>Project members of LSDIS lab projects Bioinformatics for Glycan Expression and METEOR-S (incl. Miller, Kochut, Arpinar) </li></ul><ul><li>Special thanks in background research & preparation: Karthik Gomadam, Christopher Thomas, Kunal Verma </li></ul>
    2. 3. Excellent starting point for complementary material (a partial list) <ul><li>“ Building a Bioinformatics Nation ,“ Lincoln Stein’s Keynote at O'Reilly's Bioinformatics Technology Conference 2002 </li></ul><ul><li>“ Bio-Ontologies: Their creation and design ” Peter Karp, Robert Stevens and Carole Goble </li></ul><ul><li>“ Query Processing with Description Logic Ontologies over Object-Wrapped Databases ” Martin Peim, Enrico Franconi, Norman Paton and Carole Goble </li></ul><ul><li>“ Ontologies for molecular biology and bioinformatics ” Steffen Schulze-Kremer (paper) </li></ul><ul><li>“ Can we do better than Google? Using semantics to explore large heterogeneous knowledge sources,” Anatole Gershman, SWDB Workshop, 2003. </li></ul>
    3. 4. Some current BioInformatics Systems <ul><ul><li>Tambis </li></ul></ul><ul><ul><li>BioMediator </li></ul></ul><ul><ul><li>Biodas </li></ul></ul><ul><ul><li>BioSem </li></ul></ul><ul><ul><li>Data integration, in some cases using an ontology. Single access point for multiple biological information sources; querying multiple sources. </li></ul></ul>
    4. 5. Outline of this talk… <ul><li>A Short History of Science </li></ul><ul><li>Challenges in biology </li></ul><ul><li>What can BioInformatics do for Biology? </li></ul><ul><li>What can Semantics do for BioInformatics? </li></ul><ul><ul><li>Some examples of Semantics-powered Bioinformatics </li></ul></ul>
    5. 6. What is difficult, tedious and time consuming now … What genes do we all have in common?* Research to answer this question took scientists two years** * G. Strobel and J Arnold.  Essential Eukaryotic Core, Evolution (to appear, 2003) **but we now believe with semantic techniques and technology, we can answer similar questions much faster
    6. 7. Why? Bioinformatics, ca. 2002 Bioinformatics In the XXI Century From
    7. 8. Science then , then and now In the beginning, there was thought and observation.
    8. 9. Science then , then and now For a long time this didn’t change. <ul><li>Man thought it would be enough to reason about the existing knowledge to explore everything there is to know. </li></ul><ul><li>Back then, one single person could possess all knowledge in his cultural context. </li></ul>
    9. 10. The achievements are still admirable … Reasoning and mostly passive observation were the main techniques in scientific research until recently. … as we can see
    10. 11. Science then, then and now A vast amount of information
    11. 12. Science then, then and now No single person, no group has an overview of what is known . Known, But not known …  not known
    12. 13. We don’t always know what we are looking for. <ul><li>Today’s experiments yield massive amounts of data. </li></ul><ul><li>We don’t only use the data to verify a hypothesis </li></ul><ul><ul><li>We use the data also to find new hypotheses </li></ul></ul><ul><ul><li>Computing techniques help to form hypotheses </li></ul></ul>
    13. 14. Science then, then and now
    14. 15. Science then, then and now Ontologies embody agreement among multiple parties and capture shared knowledge. Ontology is a powerful tool to help with communication, sharing and discovery. We are able to find relevant information ( semantic search/browsing ), connect knowledge and information ( semantic normalization/integration ), find relationships between pieces of knowledge from different fields ( gain insight, discover knowledge )
    15. 16. Intervention by Ontologies, some near future Bioinformatics In the XXI Century
    16. 17. Outline of the talk… <ul><li>A Short History of Science </li></ul><ul><li>Challenges in biology </li></ul><ul><li>What can BioInformatics do for Biology? </li></ul><ul><li>What can Semantics do for BioInformatics: </li></ul><ul><ul><li>Some examples of Semantics-powered Bioinformatics </li></ul></ul>
    17. 18. Challenges in biology <ul><li>What makes us ill or unwell? </li></ul><ul><ul><li>Disease identification, disease inducing agents </li></ul></ul><ul><li>What keeps us healthy and makes us live longer? </li></ul><ul><ul><li>Drug discovery </li></ul></ul><ul><li>Where do we all come from and what are we made of? </li></ul><ul><ul><li>Genetics and beyond </li></ul></ul>
    18. 19. … and their implications <ul><li>Understand biological structures of increasing complexity: </li></ul><ul><ul><li>Genes ( Genomics ): 1980s </li></ul></ul><ul><ul><li>Proteins ( Proteomics ): 1990s </li></ul></ul><ul><ul><li>Complex Carbohydrates ( Glycomics ): 2000s </li></ul></ul><ul><li>Understand biological processes and the roles structures play in them (biosynthesis and biological processes) </li></ul>
    19. 20. Outline <ul><li>Evolution of Science </li></ul><ul><li>Challenges in biology </li></ul><ul><li>What can bioInformatics do for Biology? </li></ul><ul><li>What can Semantics do for bioInformatics? </li></ul><ul><ul><li>Some examples of Semantics-powered Bioinformatics </li></ul></ul>
    20. 21. What can BioInformatics do? <ul><li>Analyze genetic and molecular sequences </li></ul><ul><ul><li>Look for patterns, similarities, matches </li></ul></ul><ul><ul><li>Identify structures </li></ul></ul><ul><li>Store derived information </li></ul><ul><ul><li>Large databases of genetic information </li></ul></ul>
    21. 22. Outline <ul><li>Evolution of Science </li></ul><ul><li>Challenges in biology </li></ul><ul><li>What can bioInformatics do for Biology? </li></ul><ul><li>What can Semantics do for bioInformatics? </li></ul><ul><ul><li>Some examples of Semantics-powered Bioinformatics </li></ul></ul>
    22. 23. Paradigm shift over time: Syntax -> Semantics <ul><li>Increasing sophistication in applying semantics & value add </li></ul><ul><li>Relevant Information (Semantic Search & Browsing) </li></ul><ul><li>Semantic Information Interoperability and Integration </li></ul><ul><li>Semantic Correlation/Association, Analysis, Insight and Discovery </li></ul>
    23. 24. Broad Scope of Semantic (Web) Technology Other dimensions: how agreements are reached, … Lots of Useful Semantic Technology (interoperability, Integration) Cf: Guarino, Gruber Gen. Purpose, Broad Based Scope of Agreement Task/ App Domain Industry Common Sense Degree of Agreement Informal Semi-Formal Formal Agreement About Data/ Info. Function Execution Qos Current Semantic Web Focus Semantic Web Processes
    24. 25. Knowledge Representation and Ontologies Catalog/ID General Logical constraints Terms/ glossary Thesauri “ narrower term” relation Formal is-a Frames (properties) Informal is-a Formal instance Value Restriction Disjointness, Inverse, part of… Ontology Dimensions After McGuinness and Finin Simple Taxonomies Expressive Ontologies Wordnet CYC RDF DAML OO DB Schema RDFS IEEE SUO OWL UMLS GO KEGG GlycO TAMBIS EcoCyc BioPAX
    25. 26. GlycO: Glycan Structure Ontology UGA’s “Bioinformatics for Glycan Expression” proj. Not just Schema/Description (partial view shown), also description base/ontology population. In progress, uses OWL.
    26. 27. What can current semantic technology do? (sample) <ul><li>Semi-automated (mostly automated) annotation of resources of various, heterogeneous sources (unstructured % , semi-structured, structured data; media content)* </li></ul><ul><li>Creation of large knowledge bases (ontology population) from the trusted sources * </li></ul><ul><li>Unified access to multiple sources*, # </li></ul><ul><li>Inferenceing # </li></ul><ul><li>Relationship/knowledge discovery among the annotated resources and entities; analytics* % </li></ul><ul><ul><li>Both implicit^ and explicit* relationships </li></ul></ul>* Commercial: Semagix; % : Near-commercial: IBM/SemTAP; # Commercial: Network Inference; ^ LSDIS-UGA Research
    27. 28. Industry Efforts (examples with bioinformatics applications only) <ul><li>Accenture’s Knowledge Discovery Tool (pre-ontology?, not product) </li></ul><ul><li>Semagix’s Semantic Browsing and Querying application for drugs for Rogers MIS and its pharmaceutical customers (product and applications); also semantic analysis application (not discussed here) </li></ul><ul><li>Network Inference’s Cerebra server: semantic engineering based bioinformatics system aiding drug discovery (product) </li></ul>
    28. 29. Existing Systems using Semantics for Bioinformatics FOCUS: SEMANTIC SEARCH AND BROWSING (with nascent work in discovery)
    29. 30. Recent Articles Experts Organizations Metabolic Pathways Protein Families Proteins Genes Related Diseases
    30. 31. Semagix Freedom Architecture (a platform for building ontology-driven information system) Ontology © Semagix, Inc. Content Sources Semi- Structured CA Content Agents Structured Unstructured Documents Reports XML/Feeds Websites Email Databases CA CA Knowledge Sources KA KS KS KA KA KS Knowledge Agents KS Metabase Semantic Enhancement Server Entity Extraction, Enhanced Metadata, Automatic Classification Semantic Query Server Ontology and Metabase Main Memory Index Metadata adapter Metadata adapter Existing Applications ECM EIP CRM
    31. 32. Practical Ontology Development Observation by Semagix <ul><li>Ontologies Semagix has designed: </li></ul><ul><li>Few classes to many tens of classes and relationships (types); very small number of designers/knowledge experts; descriptional component (schema) designed with GUI </li></ul><ul><li>Hundreds of thousands to over 10 million entities and relationships (instances/assertions/description base) </li></ul><ul><li>Few to tens of knowledge sources; populated mostly automatically by knowledge extractors </li></ul><ul><li>Primary scientific challenges faced: entity ambiguity resolution and data cleanup </li></ul><ul><li>Total effort: few person weeks </li></ul><ul><li>Key requirement: trusted knowledge sources </li></ul>
    32. 33. 1. Ontology Model Creation (Description) 2. Knowledge Agent Creation 3. Automatic aggregation of Knowledge 4. Querying the Ontology Ontology Creation and Maintenance Steps © Semagix, Inc. Ontology Semantic Query Server
    33. 35. Cerebra’s myGrid Framework
    34. 36. Outline <ul><li>Evolution of Science </li></ul><ul><li>Challenges in biology </li></ul><ul><li>What can bioInformatics do for Biology? </li></ul><ul><li>What can Semantics do for BioInformatics? </li></ul><ul><ul><li>Some examples of Semantics-powered Bioinformatics </li></ul></ul>
    35. 37. Applying Semantics to BioInformatics : Example 1 Semantic Browsing, Querying and Integration
    36. 38. Present: User queries multiple sources Heterogeneous data sources on the web ?
    37. 39. Future: the Web-Service queries multiple sources
    38. 40. Semantic Querying, Browsing, Integration to find potential antifungal drug targets Databases for different organisms Is this or similar gene in other organism? (most Antifungals are associated with Sterol mechanism ) Services: BLAST, Co-expression analysis, Phylogeny; If annotated, directly access DB, else use BLAST to normalize FGDB
    39. 41. Applying Semantics to BioInformatics : Example 2 Analytics in Drug Discovery
    40. 42. Analytics, Using Explicit and Implicit Relationships in Drug Discovery <ul><li>Some molecules contain functional groups that inhibit them from acting as drugs </li></ul><ul><li>Elimination of these groups can make the molecule, a DRUG. </li></ul>
    41. 43. Step 1: Capture domains using ontologies <ul><li>Pathogen X </li></ul>MOLECULE ONTOLOGY Molecule A Compound A Compound B DISEASE ONTOLOGY Disease D PATHOGEN ONTOLOGY Pathogen X Protein P Protein Q
    42. 44. <ul><li>Pathogen </li></ul><ul><li>X </li></ul><ul><li>Protein </li></ul><ul><li>P </li></ul>Step 2: Traverse explicit relationships <ul><li>Compound </li></ul><ul><li>A </li></ul>DISEASE ONTOLOGY Disease D PATHOGEN ONTOLOGY Pathogen X Protein Q STEP 1: 1.Look up the disease ontology 2. Identify the disease causing pathogen. STEP 2: 1.Look up the pathogen ontology 2. Identify the molecular composition of the pathogen. MOLECULE ONTOLOGY Molecule A Compound B Compound C STEP 3: 1.Look up the molecule ontology 2. Identify the composition of the possible drug.
    43. 45. Step 3: Discovering Implicit relationships… <ul><li>Protein </li></ul><ul><li>P </li></ul>MOLECULE ONTOLOGY Molecule A Compound A Compound B PATHOGEN ONTOLOGY Pathogen X Protein Q Compound A inhibits the effect of the pathogen by killing protein P Compound B produces a toxin on reacting with Protein Q Host Check if the host has protein P Extract the relationships amongst the compounds of the potential drug and the pathogen .
    44. 46. Inferences Based on Relationships <ul><li>Compound B doesn’t contribute to the curing aspect of the drug, but rather generates a toxin. </li></ul><ul><li>Eliminate compound B and molecule A can be a potential drug. </li></ul><ul><li>However if the host has protein P, then we cannot use protein to bind the drug. </li></ul><ul><li>So look for another drug that can bind at protein Q without producing a toxin </li></ul><ul><li>Eliminate and Discover!!!! </li></ul>
    45. 47. Applying Semantics to BioInformatics : Example 3 Using Ontologies in cancer research
    46. 48. Disparate Data from Different Experiments Metastatic cancer cells Increased GNT-V Activity Experiment 1 Experiment 2 Cancer marker glycan sequence elevated in glycoprotein beta 1 integrin
    47. 49. Knowledge Stored in Ontologies <ul><li>GO Ontology </li></ul><ul><ul><li>GNT V is an enzyme involved in production of N-glycans </li></ul></ul><ul><li>Glycan Structure Ontology (GlycO) </li></ul><ul><ul><li>Sequences and structures of Glycans </li></ul></ul><ul><li>Extended Structure Ontology </li></ul><ul><ul><li>How the structures are made </li></ul></ul><ul><ul><ul><li>E.g GNT V produces certain class of N-Glycans </li></ul></ul></ul>
    48. 50. Finding New Information <ul><li>Combine data from experiments and knowledge stored in ontologies </li></ul><ul><li>Known assertion from Experiments </li></ul><ul><ul><li>Beta Integrin is involved in cancer </li></ul></ul><ul><li>New assertion to be tested </li></ul><ul><ul><li>Are any other glycoproteins involved in cancer ? </li></ul></ul>
    49. 51. Applying Semantics to BioInformatics : Example 4 Applying Semantics to BioInformatics Processes
    50. 52. Creating BioSemantic Processes <ul><li>Question: What essential genes do we all have in common? </li></ul><ul><li>Research process for this using current techniques takes long time (2 years) G. Strobel and J Arnold.  Essential Eukaryotic Core, Evolution (to appear, 2003) </li></ul><ul><li>Let us demonstrate use of Functional, Data, QoS and Execution Semantics to automate the process and reduce time </li></ul>
    51. 53. Creating BioSemantic Processes <ul><li>Process for the question: </li></ul><ul><li>Input GO Id of relevant gene </li></ul><ul><li>Retrieve similar sequences </li></ul><ul><li>Perform multiple sequence alignment </li></ul><ul><li>Construct phylogenetic tree </li></ul>
    52. 54. BioSemantic Process Definition <ul><li>Create semantically annotated Web Services (wrapping tools) and process using METEOR-S </li></ul><ul><li>Semantic templates at each node allow choosing multiple tools automatically </li></ul><ul><li>Run multiple instances of process using different tools, combine results </li></ul>GO id SIMILAR SEQUENCES MATHCER MULTIPLE SEQUENCE ALIGNMENT PHYLOGEN TREE CREATOR
    53. 55. Semantic Bioinformatics Processes SIMILAR SEQUENCES MATHCER GO id MULTIPLE SEQUENCE ALIGNMENT PHYLOGEN TREE CREATOR FUNCTIONAL SEMANTICS Use Functional Ontologies for finding relevant services Sequence Matcher ENTREZ FETCH LOOKUP Sequence Alignment CLUSTAL MEME Phylogen Tree Creator PAUP PHYLIP TREEVIEW
    54. 56. Semantic Bioinformatics Processes SIMILAR SEQUENCES MATHCER GO id MULTIPLE SEQUENCE ALIGNMENT PHYLOGEN TREE CREATOR Use QoS Ontologies to make a choice Sequence Matcher ENTREZ FETCH QOS Time X secs Reliability X % QoS SEMANTICS QOS Time X secs Reliability X %
    56. 58. Semantic Bioinformatics Processes GO id CLUSTAL PAUP EXECUTION SEMANTICS Use Execution Semantics for execution monitoring of different instances GO id MEME PHYLIP GO id CLUSTAL PAUP FETCH ENTREZ ENTREZ
    57. 59. Semantic Web Process Design Template Construction
    58. 60. Common genes
    59. 61. Conclusion <ul><li>Biology research at unique standpoint in history </li></ul><ul><li>Earlier </li></ul><ul><ul><li>Biology divided into many sub-fields </li></ul></ul><ul><ul><li>Significant progress in those domains </li></ul></ul><ul><li>Present </li></ul><ul><ul><li>Searching for the bigger picture </li></ul></ul><ul><ul><li>Need to combine knowledge from sub-fields </li></ul></ul><ul><ul><ul><li>Disparate sources, terminologies </li></ul></ul></ul><ul><li>Semantics is the vehicle to get to the answers </li></ul>
    60. 62. Conclusion <ul><li>Need to capture the bioinformatics domains using ontologies </li></ul><ul><ul><li>Not just schema, instances are required </li></ul></ul><ul><li>Presented the use of semantics in </li></ul><ul><ul><li>Analytics </li></ul></ul><ul><ul><li>Knowledge discovery </li></ul></ul><ul><ul><li>Process Automation </li></ul></ul><ul><li>Collaboration between scientists and computer scientists with semantic techniques and tools make hard things easier </li></ul>
    61. 63. Sources <ul><li>Picture on each sub-section title : </li></ul><ul><li>Pictures on the title collage were taken from </li></ul><ul><ul><li> - Glycomics </li></ul></ul><ul><ul><li> - Proteomics </li></ul></ul><ul><ul><li> - Webservices </li></ul></ul><ul><ul><li> - ontology </li></ul></ul><ul><ul><li> - RNA analysis workflow </li></ul></ul><ul><ul><li> - Genomics </li></ul></ul><ul><ul><li>Molecular Modelling – Principles and Applications by Andrew R Leach - Drug Discovery </li></ul></ul>