The DrugLogics initiative is a systems medicine approach to employ computational methods for predicting drug resistance in cancer. The long-term goal is to economize drug screens, and to tailor-make treatments for patients with specific types of cancer. Currently, our pipeline can automatically generate logical models from causal statements to predict drug targets and drug synergies. In systems biology, substantial biological networks are built to depict how components in our systems are interconnected and behave. A considerable amount of knowledge provided by different public resources is available in the form of large biological networks. We are exploring ways to make this public information more easily computationally available for regulatory network building by breaking down the networks into their most basic regulatory network motifs: causal statements, a source entity having an influence over the quantity or the activity of a target entity.
Two specific projects within the DrugLogics initiative focus on:
1) the extraction of causal statements from existing network resources to feed our model building software pipeline with ample amounts of data. We will design a format that allows to standardize the representation of causal statements by using generally accepted identifiers and ontology terms (i.e. ontology terms for causal relationships). Using this representation format we will design and build software pipelines to extract causal statements from a variety of existing network resources and make them publicly available.
2) the standardisation of the boolean models generated from the pipeline to comply with SBMLqual and SBGN in order to facilitate their use by the community.
This project is in collaboration with teams at EBI, IBENS and Institut Curie.
Links to projects:
https://www.ntnu.edu/health/druglogics
http://www.colosys.org
2. The DrugLogics project
DrugLogics
Crossover Research
Structured Knowledge
Commons resource
DbTF curation
Scicura
COLOSYS
Drug resistance
prediction in colon cancer
via computer models
Drug Combinations
Development of anti-
cancer combinations
Towards the development of precision and personalised medicine
http://www.druglogics-ntnu.org, soon: https://www.DrugLogics-EU.org
3. My tasks within the DrugLogics/NTNU-
Health project
prior knowledge
DrugLogics’ pipeline
“CausalDB” Logical modeling to predict
drug targets and synergies
Representation of
causal statements
Extraction from
prior knowledge
Availability of
the data
A->B
C-| D
Models
Facilitate the process of building biological models with causal statements
4. Representation of causal statements
AKT1 FOXO3
Down-regulates activity of
Causal interaction between two biological entities (gene, RNA, protein,
complexes, etc…)
5. Representation of causal statements
Is it a direct or indirect
Interaction?
Which molecular function
is down-regulated?
What is FOXO3’s state?
(active/inactive)
What is the regulation type?
(phosphorylation, acetylation,
dephosphorylation)
How to represent meaningful causal interactions?
When and where does
this interaction occurs?
AKT1 FOXO3
Down-regulates activity of
6. Representation of causal statements
Entity – Source (Regulator) / Target (Regulated)
● ID – ex:causalDB:FOXO3
● Reference ID - HGNC; Uniprot; Entrez
● For Complex: ComplexPortal ID?
● For Families?
● Name – ex:FOXO3
● Molecule type - gene, RNA, protein, complex
● Acting entity – Reference ID
● Molecular function – GO:MF
● State – active / inactive
Research ideas on information that should be ideally encoded
Causal Relationship
● Regulation type – down-regulates
● Mechanism - PSI MOD?
● Modified residue – Tyr@P202
● Interaction depth – 0 (direct); 1; 2; etc…
● Text – Scicura (http://scicura.org/info.html)
● Provenance – ex:Reactome
● Evidence – ECO
● Confidence score?
Controlled Vocabulary and Ontologies – essential to make data sustainable, shareable and interoperable
Context
● Species – TaxID
● Tissue type – Brenda Tissue Ontology (BTO), Uberon?
● Cell type – BTO, Cell Line Ontology (CLO)?
● Experimental conditions – if evidence is experimental
● Tissue / Cell state
Causal
Relationship
EntityContext
hasSource
hasTarget
hasContext
hasContext?
7. Representation of causal statements
Controlled vocabulary / ontologies for representing causal interaction type
Effect Gene
Ontology
PSI-MI causal
interaction
Relation
Ontology
BEL
statement
IntAct Signalink
Positive
regulation
positively
regulates
up-regulates activity directly
positively regulates
activity of
increases activates stimulation
Negative
regulation
negatively
regulates
down-regulates activity directly
negatively regulates
activity of
decreases inhibits inhibition
→ Need unification
8. Extraction from prior knowledge
DB of causal
interactions
DB of molecular
interactions
Pathways,
reactions
Pathways of
cancer related
signaling
networks
Aggregation of causal data from several existing resources
9. Example: extraction from Reactome
ReactionLikeEvent
PhysicalEntity
CatalystActivityRegulation
ReferenceEntity
catalystActivityregulatedBy
regulator physicalEntity
outputinput
referenceEntity
Reactome data model extraction using Neo4j and Cypher Query language
10. Example: extraction from Reactome
“Cypher is your friend” - A. Fabregat
MATCH (rle:ReactionLikeEvent)-[:regulatedBy|catalystActivity]->(o)-[:regulator|physicalEntity]->(source:PhysicalEntity)
OPTIONAL MATCH (input:PhysicalEntity)<-[:input]-(rle)-[:output]->(output:PhysicalEntity)
RETURN rle.stId AS ReactionID,
rle.displayName AS Reaction,
COLLECT(input.displayName) AS Inputs,
COLLECT(output.displayName) AS Outputs,
o.simpleLabel AS Regulation,
source.displayName AS Regulator
Example: Get all reactions regulated by a physical entity or catalysed by a catalyst activity
Reactome data model extraction using Neo4j and Cypher Query language
11. Example: Resulting outputs
ReactionId Reaction Compartment Inputs Effect Outputs Regulator
R-HSA-
452338
Expression of TDGF1
(CRIPTO)
cytosol ["TDGF1 gene
[nucleoplasm]"]
NegativeGeneExp
ressionRegulation
["N-aspartyl-
glycosylphosphatidyli
nositolethanolamine-
TDGF1(31-188)
[plasma
membrane]"]
NR6A1(GCNF):TDGF1
gene [nucleoplasm]
R-HSA-
8936628
GP1BA gene
transcription is
stimulated by the
complex containing
RUNX1, PRMT1 and
GATA1 and inhibited
by the complex of
RUNX1, SIN3A and
PRMT6
plasma
membrane
["GP1BA gene
[nucleoplasm]"]
NegativeGeneExp
ressionRegulation
["GP1BA [plasma
membrane]"]
RUNX1:CBFB:SIN3A,
(SIN3B):PRMT6:HDA
C1:GP1BA
gene:H3K4me2,H3R2
me2a-Nucleosome
[nucleoplasm]
R-HSA-
8944497
PTEN mRNA
translation is
negatively regulated
by microRNAs
cytosol ["PTEN mRNA
[cytosol]"]
NegativeGeneExp
ressionRegulation
["PTEN [cytosol]"] miR-20 RISC:PTEN
mRNA [cytosol]
Causal
interaction
SourceTarget?
12. Example: questions / issues raised
● Exclude trivial molecules
● Missing IDs for the modified mechanism type
● How to transform a reaction network to causal statements?
● Definition of necessary and sufficient contextual information
→ MICAST: Minimum Information for representing Causality Statements?
malate + NAD+ <=> oxaloacetate + NADH + H+
A’A
C C -| A
C -> A’
A -| A’
A’ -| A
13. Current work
● Collaboration with Curie and ENS - Paris
– Define the representation of causal statements
– Extraction of causal statements from ACSN
– Consensus representations with GO, PSI-MI,
COLOMOTO
– Standardisation of the pipeline → SBML-Qual?
14. Thank you for your attention!
Rune Nydal
Ane Møller Gabrielsen
Anamika Chatterjee
Martin Kuiper
Steven Vercruysse
Wim De Mulder
Vladimir Mironov
Vasundra Touré
Stian Holmås
Rafel Riudavets
Astrid Lægreid
Liv Thommesen
Åsmund Flobak
Marcio Acencio
Barbara Niederdorfer
Evelina Folkesson
Kathleen Heck
Funding:
NTNU Helse
Dept of Biology Dept of Clinical and
Molecular Medicine
Dept of Philosophy
and Religious Studies
The DrugLogics team