Increasingly sophisticated knowledge about RNA structure and function requires an inclusive knowledge representation that facilitates the integration of independently-generated information arising from such efforts as genome sequencing projects, microarray analyses, structure determination and RNA SELEX experiments. While RNAML, an XML-based representation, has been proposed as an exchange format for a select subset of information, it lacks machine-understandable semantics that make it arbitrarily user-extensible, as is the case for formal logic based languages. Here, we describe an RNA knowledge base (RKB) for structure-based knowledge using RDF/OWL Semantic Web technologies. RKB contains basic terminology for nucleic acid composi-tion along with context/model-specific representation of structural features such as sugar conformations, base pairings and base stackings. RKB is populated with RNA PDB entries and MC-Annotate structural annotation. The use of semantic web technologies addresses the reality of diverse interests of the RNA Ontology Consortium and supports knowledge discovery over independently-published RNA knowledge.

1. RKB – A Semantic Knowledge Base for RNA
Michel Dumontier 1, José Cruz-Toledo 1
Marc Parisien 2, Francois Major 2
1 Carleton University
2 Université de Montreal

2. Objectives
i. To represent biochemistry of nucleic
acids and their structural characteristics
including base pairing/stacking
ii. Represent context specific knowledge
iii. Capture the structural annotation
generated by MC-Annotate
5/25/2009 Carleton University -- Dumontier Lab dumontierlab.com 2

3. Guided design
• Modeling with Upper Level Ontologies
– interoperability and semantic coherency
– New Upper Level Ontology (NULO)
• distinguishes objects, qualities, roles, processes
and spatial regions
• Based on BFO/RO, but for OWL
5/25/2009 Carleton University -- Dumontier Lab dumontierlab.com 3

4. Biological Modeling
• Objects
– Occupy space
• Nucleic acids, nucleotides, riboses and phosphates
• Qualities
– Intrinsic categorical or numeric valued property
• Nucleotide bears the quality of conformation
• Roles
– Defined by extrinsic interactions
• A C3’ atom may hold the exo role during some sugar puckering
• Processes
– Entities that extend in time
• structure determination, an interaction
5/25/2009 Carleton University -- Dumontier Lab dumontierlab.com 4

5. Contextual Modeling of Nucleic Acids
• Base stacking varies in different XRD/NMR models
• Need to know in which model that info is found
• We want to set the stage for representing simulation.
5/25/2009 Carleton University -- Dumontier Lab dumontierlab.com 5

6. RKB populated with PDB, MC-Annotate
• The ontology population involved 3 steps:
i. Assigning names
ii. Asserting class membership
iii. Assigning relations between entities
• The following naming convention was used:
– Objects:
• Polymer: PDBID_cCHAIN
• Residue: PDBID_cCHAIN_rRESIDUE
• Atom: PDBID_cCHAIN_rRESIDUE_aAtom
– Quality/Roles
• PDBID_mMODEL_cCHAIN_rRESIDUE_type
– Processes
• Structure determination: PDBID_mMODEL
• Interaction: PDBID_mMODEL_PROCESSTYPE_PARTICIPANT
5/25/2009 Carleton University -- Dumontier Lab dumontierlab.com 6

7. Support for Leontis-Westhof Nomenclature
• The RKB incorporates LW nomenclature
• Describes the three edges for H-bonding
interactions in purines (Y) and pyrimidines (R)
• Atom composition:
i. Watson-Crick Edge:
• A(N6)/G(O6), R(N1), A(C2)/G(N2), U(O
4)/C(N4), Y(N3) and Y(O2)
ii. Hoogsteen Edge (CH edge for R):
• A(N6)/G(O6), R(N7), U(O4)/C(N4) and
Y(C5)
iii. Sugar Edge:
• A(C2)/G(N2), R(N3), Y(O2) and O2’
• cis and trans orientations
• relative orientations of the glycosidic bond
between the sugar and the PO4 group
5/25/2009 Carleton University -- Dumontier Lab dumontierlab.com 7

8. Support for LW+ Nomenclature
• Extension incorporates faces to each edge:
– WC edge:
• Wh, Ww and Ws faces
– Hoogsteen Edge:
• C8(Y), Hh, Hw and Bh
– Sugar Edge:
• Bs, Ss(Y), Sw and O2’
• The Bh and Bs faces involve the Hoogsteen
side amino/keto group and the sugar side
amino/ keto group respectively.
• The C8 face was introduced for the C8-H8
donor group in purines
5/25/2009 Carleton University -- Dumontier Lab dumontierlab.com 8

9. Describing Base Pairs
• Base pairs composed of
interactions with the edges or faces
of the interacting bases
• Role chains capture additional
knowledge:
Objects that participate in sub-processes (face
interactions) are also participants of the
process whole (base pair)
hasPart ◦ hasParticipant -> hasParticipant
Objects are involved in processes when their
qualities are
isBearerOf ◦ isParticipantIn -> isParticipantIn
5/25/2009 Carleton University -- Dumontier Lab dumontierlab.com 9

10. The RKB is compatible with both the LW and
the Saenger nomenclature for base pairs
• The semantics of the RKB enables
the usage of consistent bp naming
schemes
• The AA BP in model 4 of
PDB:1B36 can be classified as the
being member of the following
classes:
– Saenger type II A A
– LW Trans Hoogsteen/Hoogsteen (8)
NucleotideBasePair
and ParallelBasePair
and TransBasePair
and HoogsteenHoogsteenBasePair
and hasAgent exactly 2 AMP
Carleton University :: Dumontier Lab ::
5/25/2009 10
dumontierlab.com

11. Sugar Puckering
• The ribose ring presents two
distinct puckering
modes, envelope and twist
• The classification into either
geometry is dependent on the
relative position of the carbon
atoms of the ribose to its C5’
atom
• Carbon atoms in a ribose thus
bear either the endo or exo role
with respect to the plane
formed by the other atoms
5/25/2009 Carleton University -- Dumontier Lab dumontierlab.com 11

12. Sugar Puckering (cont’d)
Our implementation of situational
modeling assures that objects are
represented by a single entity
throughout their lifetime, thus
avoiding the need to create
multiple distinct instances of the
same object in each particular
spatial-temporal context with
different attributes
5/25/2009 Carleton University -- Dumontier Lab dumontierlab.com 12

13. RKB is SPARQL accessible
• SPARQL is a graph query language
• Loaded instantiated ontology into Virtuoso 6
• SPARQL endpoint
– http://codemonkey.dumontierlab.com/sparql/
• Specify Graphs to restrict search
– http://semanticscience.org/rkb/mcannotate/pdb/dna
– http://semanticscience.org/rkb/mcannotate/pdb/rna
Carleton University :: Dumontier Lab ::
5/25/2009 13
dumontierlab.com

14. Query 1: Find all face interactions
(model 1 of PDB:1B36)
PREFIX ss: <http://semanticscience.org/>
select distinct ?faceInteraction
where {
?pair ss:isProperPartOf <http://semanticscience.org/pdb/1B36_m1> .
?pair ss:hasProperPart ?faceInteraction .
?faceInteraction rdf:type ss:FaceInteraction .
}
Nucleotide base pairs are composed of one or more face interactions.
Where known, such as in the MC-Annotate results, we can retrieve all 18
instances of this that satisfy this query.
Carleton University :: Dumontier Lab ::
5/25/2009 14
dumontierlab.com

15. See results : http://tinyurl.com/porxdb
5/25/2009 Carleton University -- Dumontier Lab dumontierlab.com 15

16. Query 2: Find all C8 mediated base pairs
(model 1 of PDB:1B36)
PREFIX ss: <http://semanticscience.org/>
SELECT DISTINCT ?faceInteraction ?residue ?hasC8Face
where {
?pair ss:isProperPartOf <http://semanticscience.org/pdb/1B36_m1> .
?pair ss:hasProperPart ?faceInteraction .
?faceInteraction rdf:type ss:FaceInteraction .
?C8Face ss:isAgentIn ?faceInteraction .
?C8Face rdf:type ss:C8Face .
?residue ss:hasQuality ?C8Face
}
Results: http://tinyurl.com/r7b5e4
Face interactions are mediated by the faces of bases. Nucleotides and
their face qualities are related by the hasQuality relation, whereas faces
are agents in the face interaction, and are related by the hasAgent
relation.
Carleton University :: Dumontier Lab ::
5/25/2009 16
dumontierlab.com

17. Query 3: Find base pairs involving a GMP sugar-sugar face
(model 1 of PDB:1B36)
PREFIX ss: <http://semanticscience.org/>
SELECT distinct ?faceInteraction ?residue ?hasSSFace
WHERE {
?pair ss:isProperPartOf <http://semanticscience.org/pdb/1B36_m1> .
?pair ss:hasProperPart ?faceInteraction .
?faceInteraction rdf:type ss:FaceInteraction .
?hasSSFace rdf:type ss:SugarSugarFace .
?hasSSFace ss:isAgentIn ?faceInteraction .
?residue ss:hasQuality ?hasSSFace .
?residue rdf:type ss:GMP
}
Results found at: http://tinyurl.com/qpup8z
This query builds on Query 2, in that it requires a Ss face to be on an
AMP that is participating in a base pair.
Two GMPs are found to have this particular face participating with other
nucleotides in base pairs in this particular structure
Carleton University :: Dumontier Lab ::
5/25/2009 17
dumontierlab.com

18. Query 4: Find Hoogsteen – O2’ face interactions
(model 1 of PDB:1B36)
PREFIX ss: <http://semanticscience.org/>
SELECT distinct ?faceInteraction ?residue1 ?residue2 ?hasHhFace ?hasO2pFace
where {
?pair ss:isProperPartOf <http://semanticscience.org/pdb/1B36_m1> .
?pair ss:hasProperPart ?faceInteraction .
?faceInteraction rdf:type ss:FaceInteraction .
?hasHhFace rdf:type ss:HoogsteenHoogsteenFace .
?hasHhFace ss:isAgentIn ?faceInteraction .
?hasO2pFace rdf:type ss:O2pFace .
?hasO2pFace ss:isAgentIn ?faceInteraction .
?residue1 ss:hasQuality ?hasHhFace .
?residue2 ss:hasQuality ?hasO2pFace
}
Results found at: http://tinyurl.com/oo4fp8
LW+ nomenclature more detailed for base interactions.
The result of this query describes a single base pair in this structure.
Carleton University :: Dumontier Lab ::
5/25/2009 18
dumontierlab.com

19. Future Directions
• Specify Saenger nomenclature
• Map other structural annotator output (e.g. 3DNA)
• Extend structural knowledge with 6 backbone angles
– range restrictions on classes
• SWRL / DL-safe rules or SPARQL query required to
specify cyclic motifs
• Publish as part of Bio2RDF network
5/25/2009 Carleton University -- Dumontier Lab dumontierlab.com 19

20. RKB Availability
• Creative Commons License.
• Google Code Project:
– http://semanticscience.org
• Instructions:
http://code.google.com/p/semanticscience/wiki/RKBDownload
5/25/2009 Carleton University -- Dumontier Lab dumontierlab.com 20

21. References
• Dumontier, M., et al. (2009). RKB: A Semantic Web Knowledge
Base for RNA, Accepted in Bio-Ontologies
2009, Stockholm, Sweden
• Smith, B., et al. (2005). Relations in biomedical ontologies. Genome
Biol, 6(5): p. R46
• Leontis, N. B. and E. Westhof (2001). Geometric nomenclature and
classification of RNA base pairs. RNA, 7(4): 499-512.
• Lemieux, S. and F. Major. (2002). RNA canonical and non-canonical
base pairing types: a recognition method and complete repertoire.
Nucleic Acids Res, 30(19): p. 4250-63.
• Major, F., Thibault, P., Computer Modeling of RNA Three-
Dimensional Structures, in Encyclopedia of Molecular Cell Biology
and Molecular Medicine, R.A. Meyers, Editor. 2005, Wiley-VCH
Verlag GmbH & Co.: Weinheim. p. 605-636.
5/25/2009 Carleton University -- Dumontier Lab dumontierlab.com 21