Bibliological data science and drug discovery

Bibliological data science and drug discoveryKnowing the knowns*

Effectively Harnessing the World’s Literature To Inform Rational Compound Design - ACS National Meeting, Philadelphia, Aug 21-24, 2016

Jeremy J Yang

Translational Informatics Division School of Medicine

University of New Mexico

Integrative Data Science Lab School of Informatics & Computing

Indiana University

*phrase borrowed from Edgar Jacoby, Janssen.

In science, luck favors the prepared.- Louis Pasteur

The main thing was not to . . . "foul up." - The Right Stuff, by Tom Wolfe, about John Glenn.

Overview of talk

● Formulation of problem● Resources and examples:

TIN-X, Target Importance and Novelty Explorer (&IDG)

Chem2Bio2RDF

OPDDR, Open Phenotypic Drug Discovery Resource

DrugCentral

Formulation of problem

● "World's Literature" redefined by online revolution● Rational Compound Design = improving our odds● For given research question, what are the known knowns?● Connect the dots and weigh the evidence from global

knowledge graph.

TIN-X

TIN-X Target Importance & Novelty Explorer

● Bibliometric application developed for Illuminating the Druggable Genome (IDG) project

● Text mining from Novo Nordisk Center for Protein Research (U. Copenhagen) lab of Lars Juhl Jensen.

● Algorithm and client developed at UNM (Cristian Bologa, Daniel Cannon)

● Disease Ontology (DO) classification ● Drug Target Ontology (DTO) protein classification

Illuminating the Druggable Genome (IDG)

7Knowledge Mgmt Center PI:

Tudor Oprea, MD, PhD

pharos.nih.gov

TIN-X

http://newdrugtargets.org

TIN-X

TIN-X

http://newdrugtargets.org

Target Novelty:

Fk = 1 / Tk

● Tk = # targets in paper (k)● Fk = fractional score of paper (k)● for papers where Tk > 0

Ni = 1 / ∑(Fk)● Ni = novelty, target (i)● sum over papers where target (i) mentioned

Target-Disease Importance:

Fk = 1 / (Tk * Dk)● Tk = # targets in paper (k)● Dk = # diseases in paper (k)● Fk = fractional score of paper (k)

Iij = ∑(Fk)● Iij = importance, target (i) for disease (j)● sum over papers where both mentioned

Target Importance and Novelty Explorer (TIN-X), Daniel Cannon, Jeremy Yang, Stephen Mathias, Oleg Ursu, Subramani Mani, Anna Waller, Stephan Schürer, Lars Juhl Jensen, Larry Sklar, Cristian Bologa, and Tudor Oprea (manuscript in preparation).

TIN-X

TIN-X Target Importance & Novelty Explorer

● Text mining is a valuable tool for monitoring literature, filtering and ranking, and detecting trends.

● Automation can infer patterns regarding community trends and consensus.

● Interactive visualization tools help navigate big data.● Good big data text miners care about small data too!

TIN-X Key contributors

Cristian Bologa Daniel Cannon Lars Juhl Jensen

Chem2Bio2RDF

● 24 sources, 52 datasets, 78M triples

● Semantically linked● Chen, B, et al, BMC

Bioinformatics (2010).● Chen, B et al, PLoS

Comp Bio (2012).● Fu, G et al, BMC

Bioinfo (2016).● Related projects:

Bio2RDF, LOD

http://chem2bio2rdf.org

Classes:biological chemical

chemogenomicsliterature

phenotypesystemsdiseasepathway

polypharmacologyPPI

side effect

BindingDBBindingMOADIUChEBIChEMBLCTDDCDBDIP

DrugBankHGNCHPRDKEGGMATADOROMIMPDBePDSP

PharmGKBPubChemPubMedReactomeSIDERTTDUniProt

Sources:

Linked Open Data (LOD)

http://linkeddata.org/

Chem2Bio2RDF apps: (1) SLAP, (2) Metapaths

2012

2016

● Data semantics essential for integration of heterogeneous sources

● Strong evidence requires strong semantics● Semantic Web Technologies common framework

enabling -- but not assuring -- community progress● Chem2Bio2RDF v2.0 to leverage major community

advances (esp. Open PHACTS)● Data ecosystems, coop-tition & prisoner's dilemma

Key contributors

Bin Chen Ying Ding David Wild

OPDDR

OPDDR

Open Phenotypic Drug Discovery Resource

https://ncats.nih.gov/expertise/preclinical/pd2

OPDDRcollaboration

Example: OIDD HeLa cell based assayIntegrated RDF

bioassay:AID1117350skos:exactMatchoidd_assay:17 .

bioassay:AID1117350 dcterms:source source:ID846 ; dcterms:title "Increased chromatin condensation in HeLa cells-IC50"@en .

bioassay:AID1117350 rdf:type bao:BAO_0002786 .bioassay:AID1117350 rdf:type bao:BAO_0000010 .bioassay:AID1117350 rdf:type bao:BAO_0000219 .

endpoint:SID170464897_AID1117349 vocabulary:PubChemAssayOutcome vocabulary:active ; sio:has-value "0.0656"^^xsd:float ; a bao:BAO_0000190 ; rdfs:label "IC50"@en .

substance:SID170464897skos:exactMatchchembl_molecule:CHEMBL1483 .

chembl_assay:OIDD00017cco:hasCellLinechembl_cell_line:CHEMBL3308376 .

D2D builds apps, tools and solutions

for knowledge discovery powered

by fast, scalable network analytics

and rigorous semantics.

d2discovery.com

Predictive Phenotypic Profiler (P3) prototype

openphacts.org

OPDDR

● OPDDR phenotypic assays have been linked and integrated via community semantics to both phenotypic (cell lines) and molecular (genomic/protein targets)

● New phenotypic knowledge domain offers additional value in drug discovery and pharmacological informatics

● Open PHACTS excellent, well suited platform

DrugCentral

DrugCentral

● DrugCentral is a free, open, curated resource about approved drugs, designed for research

● Compounds, products, labels, targets, IDs, names● DrugCentral developed over several years at UNM● DrugCentral recently released with new interface● License: CC-BY-SA

http://drugcentral.org



DrugCentral

● Free, open, accurate, comprehensive drug reference for biomolecular and biomedical informatics research

Compounds 4444

Products 84787

Synonyms 20522

Structures 4231

Targets 3651

Bioactivities 15620

MoA 3484

SNOMED 45349

"DrugCentral: online drug compendium", Oleg Ursu, Jayme Holmes, Jeffrey Knockel, Cristian Bologa, Jeremy Yang, Stephen Mathias, Stuart Nelson, Tudor Oprea (manuscript submitted).

In Conclusion● New resources continue to emerge and evolve, providing

opportunities for knowledge driven drug discovery● Community standards → more intelligent web● Adapt to new data environment for success● Private + public data must be integrated to

○ Be prepared (like Pasteur)○ Not "foul up" (like Glenn)