250 likes | 382 Views
OpenBEL Initiative BEL-RDF. Objectives. Create an RDF model that can represent BEL scripts with high (sufficient) accuracy: round-trip conversions Define a semantic mapping with as much clarity and fidelity as possible between BEL causal scripts and standard RDF semantics
E N D
Objectives • Create an RDF model that can represent BEL scripts with high (sufficient) accuracy: round-trip conversions • Define a semantic mapping with as much clarity and fidelity as possible between BEL causal scripts and standard RDF semantics • Entities, Abundances, Processes, Modifiers, Citations , Metadata • Apply various Semantic tools (SPARQL) to BEL RDF and see what is possible
Some Bio Basics… • Gene expression -> protein synthesis
Some Bio Basics… • And then proteins get modified…
Some Bio Basics… • Everything works together in…
Relevance to Drug R&D • The current model of applying chemistry and biology (reductionist assays) is not working for Pharmaceuticals companies • Drug researchers need to understand biology more completely • genomics, protein interactions, systems, cells… • That means understanding “cause and effect”!
Selventa’s OpenBEL • OpenBEL is about representing molecular and processes as cause-and-effect… proteinAbundance(MGI:Jak1,proteinModification(P)) directlyIncreaseskinaseActivity(proteinAbundance(MGI:Jak1)) proteinAbundance(MGI:Il13) increases proteinAbundance(MGI:Pik3r1,proteinModification(P)) proteinAbundance(MGI:Il13) increases proteinAbundance(MGI:Ptpn11,proteinModification(P)) proteinAbundance(MGI:Il4) increases proteinAbundance(MGI:Nr0b2,proteinModification(P)) proteinAbundance(MGI:Inpp5d) decreases transcriptionalActivity(proteinAbundance(MGI:Stat3)) proteinAbundance(MGI:Pias3) decreases transcriptionalActivity(proteinAbundance(MGI:Stat3)) complexAbundance(proteinAbundance(MGI:Stat3),proteinAbundance(MGI:Stat3)) directlyIncreasestranscriptionalActivity(proteinAbundance(MGI:Stat3)) proteinAbundance(HGNC:IL4) increases proteinAbundance(HGNC:IRS2,proteinModification(P,Y)) …
BEL Semantics • IFNA1 -> CISH transcription p(HGNC:IFNA1) r(HGNC:CISH) • :IFNA1-prot-abundance :increase_indirecteffects :CISH-mRNA-abundance . • IL2 Modifications p(HGNC:IL2, pmod(P, S, 21), trunc(33)) => r(HGNC:CISH) • :IL2-prot-abundance :has_modificationpmod(P, S, 21) , trunc(33)) ;:increase_directeffects:CISH-mRNA-abundance .
Insights into BEL • Abundances and Activities (of genes, proteins, RNA, Cmpds) are the primary actors in BEL • Abundances of some proteins can have associated Activities • Proteins can have Modifications, therefore their Modified forms also have Abundances • Abundances or Activities can influence other Abundances or Activities through Effects • Effects can be either increasing or decreasing, and either direct or indirect
RDF Semantic Mappings Some conscious decisions: • Abundances and Activities are distinct from Genes and Proteins, but associated to them as classes • Abundances and Activities can have Modifications (e.g., protein) • Effects can be modeled as Binary Operations • Effects can cause other Effects: Series and Cascades
Abundances/Activities Semantics Why are Abundances different from Molecules? • Molecules are structural entities, but in order to have physiological and biochemical effects in a particular context, additional concepts need to be included: • abundance/quantitation (Q) • activity/activation (A) • These introduce additional dimensions of quantity en mass within a locality, context, state, and dynamics • Modifications requires their own Abundances
Simple Example BEL p(HGNC:IFNA1) -> r(HGNC:CISH) // protein abundance of IFNA1 increases the mRNA of CISH RDF ns1:IFNA1_proteinAbundance a bel:ProteinAQC ; bel:affects ns1:IFNA1_CISH_IndirectIncreaseEffect_1 ; bel:has_child HGNC:IFNA1 ; bel:child_for_protein_abundance HGNC:IFNA1 . ns1:CISH_rnaAbundance a bel:MrnaAQC; bel:is_affected_byns1:IFNA1_CISH_IndirectIncreaseEffect_1 ; bel:has_childHGNC:CISH ; bel:child_for_mrna_abundance HGNC:CISH . ns1:IFNA1_CISH_IndirectIncreaseEffect_1 a bel:IndirectIncreaseEffect ; bel:rel_affects ns1:CISH_rnaAbundance ; bel:rel_is_affected_by ns1:IFNA1_proteinAbundance . Causal effect ➜ p(IFNA1) p(CISH) abundance molecule IFNA1 CISH
A little more complexRegulating regulation composite(a(CHEBI:"deoxyribonucleic acid"), a(CHEBI:"NAD(+)")) -> (ribo(p(HGNC:PARP1)) => p(HGNC:PARP1, pmod(R))) composite(a(CHEBI:"deoxyribonucleic acid"), a(CHEBI:"NAD(+)")) -> (ribo(p(HGNC:PARP1)) => p(HGNC:XRCC5, pmod(R))) p(HGNC:XRCC5, pmod(R)) => ribo(p(HGNC:PARP1)) DNA-NAD+ ↓ Evidence A ↓ NAD+ DNA ⇒+ +⟸ PARP1ribo mod PARP1ribo activity XRCC5 ribo mod XRCC5 PARP1abundance ⇒+ PARP1
BEL Ontology BEL Ontology has three super classes: Causal Nodes Script (Utility) Nodes Molecular (External) Entities
BEL Causal Classes BEL Causal classes define AQC and AQM classes, which are the Affecters and Effectors. Relationships connect them causally to each other. Modifications allow modifier extensions to the AQM classes.
BEL AQM ClassesActivity and Quantitative Molecular Classes AQM classes include the full range of molecular interaction entities. Both Quantity and Activity are considered here.
BEL Relationships ClassesCause and Correlation Relationships have to higher forms: Causal and Correlation Causes can be direct or indirect, increase or decrease. Correlations can be positive or negative.
BEL Modification ClassesState changes Modifications alter protein structures, yet each change needs to be associated with the abundance and activity of that modification type.
BEL Activities and Processes Activities are associated with dynamics and activity states.
BEL Utility and External Classes BEL ScriptNodes support all of the BEL document definitions and metadada. • Citations • Evidence • Context • Annotations • Namespace Comments (can be overridden)
Applications • Inferring downstream consequences • Finding all upstream antecedents • Indexing all +/- effectors and their dependencies • Identifying regulatory or dependence “hubs” • Comparing different observed contexts for variations in outcomes • Determining necessary conditions for certain reactions: missing factors • Hypothesis Formation
Comparisons • BioPAX – Molecular reactions ontology that currently does not support causal effects (e.g, equilibrium direction?) or bio-context • SBML – Focus on “assumed” math models; requires detailed information • OpenBEL RDF includes • Negative (absence of effect) assertions • Cataloging different contexts • Protein modification effects (switches) • New causal rules can be added as BEL statements
Summary • BEL-RDF is based on the BEL Language • BEL-RDF models effects and relations as nodes • Offers a context/evidence based view of molecular (causal) mechanisms • Intended to map external molecular and process URIs • Supports the layering of causal rules