Structure based auto-classification using ChEBI ontology
Download as PPT, PDF1 like337 views
This document discusses using an automated structure-based classification approach for the ChEBI ontology. It proposes using schema definitions that describe structural properties like having a ketone or acetate part to classify molecules. Previously, classification required manual definition generation, but this approach uses ranked schema definitions to classify molecules automatically. Challenges remain for some complex classes, but planned integration into ChEBI tools could improve data loading and curation.
Structure based auto-classification using ChEBI ontology
- 1. AUTOMATED STRUCTURE-BASED CLASSIFICATION USING ChEBI ONTOLOGY Venkatesh Muthukrishnan Software Engineer - ChEBI
- 5. BLOCKS BULK LOADING … http://sourceforge.net/p/chebi/news/2012/11/chebi-release-97-is-now-available/
- 7. PREVIOUS APPROACHES SOCO SELF-ORGANISING CHEMICAL ONTOLOGIES SMARTS & OWL Chepelev et al. BMC Bioinformatics 2012 13:3 doi:10.1186/1471-2105-13-3
- 9. SCHEMA DEFINITIONS eg: ketone organic_molecular_entity and has_part some acetone RANKED
- 10. DEFINITIONS dicarboxylic acid dianion organic_molecular_entity and has_part exactly 2 acetate and has_charge value "-2"^int flavonoid organic_molecular_entity and has_skeleton some flavones benzoquinones organic_molecular_entity and ( has_part some 1,2-benzoquinone or has_part some 1,4- benzoquinone ) gamma-lactam organic_molecular_entity and has_part some pyrrolidin-2-one and not ( has_part some succinimide )
- 15. PLANNED INTEGRATION INTERNAL DATA LOADING SUBMISSION TOOL & CURATOR TOOL
- 16. ACKNOWLEDGEMENT S COLLABORATORS Colin Batchelor, RSC Lian Duan, ETH Leonid Chepelev, Ottawa Michel Dumontier, Stanford Despoina Magka, Oxford FUNDING BBSRC “Continued development of ChEBI towards better usability for the systems biology and metabolic modelling communities” BB/K019783/1
- 17. THANKYOU
Editor's Notes
- #2: Chebi ontology has 3 sub-ontologies. Namely, role, subatomic particle and chemical ontology In this talk, I will be focusing only on the chemical ontology. An ontology that captures the structural features hierarchically.
- #3: This is an example entry for a structural classification. By looking at the graph from top to bottom we can describe few structural features for caffeine. It is certain that caffeine has at least two cycles as polycyclic compound. narrowing down that it contains only two cycles & hetero atoms as hetero bicyclic compound. An imidazopyrimidine - a 6 ring containing two nitrogens fused to 5 ring containing two nitrogens methylxanthine - imidazopyrimidine with two ketones in 6 membraned ring.
- #4: what are the challenges that are expected with manual classification of structures ?
- #7: As a result we a need a auto-classification tool that would help us to identify and correct these consistencies of ontology. And also allows us to bulk load of structures. As result we could speed up the curation process and make an consistent ontology.
- #8: SMiles ARbitrary Target Specification (SMARTS) Web ontology language (OWL) Fragmentation based approach where it captures the structural features hierarchically in SMARTS and uses owl to classify No support for negation Only “min” counting supported, not max or exactly. Thus, a dicarboxylic acid is a monocarboxylic acid SMARTS is powerful – but not very human-readable notations. Can we do better at making definitions accessible?
- #9: So the new proposed approach is to make this definitions human friendly. So any chemically intelligent person can validate this definitions without proper computer knowledge.
- #10: In this approach, the structural features are encoded in the owl definitions. As in this example we say a basic functional group ketone contains a structure of acetone. These owl definitions are parsed and converted in to chemoinformatics definitions. That are matched against the unclassified structures. As a result the structure is classified under highly ranked structural features.
- #12: These definitions are manually generated to make it more sensible. As an initial exercise MCS was used to extract the structural features to generate definitions.
- #13: In this example class benzoquinone, we have two different substituents and one is more dominant. This is the mcs result for benzoquinones by RDkit & SMSD. This makes the automatic definition generation tricky when there is multiple definitions because of substituents or ring size and so on.