an architecture for an open science molecular compound database

Department of Bioinformatics - BiGCaT 1

An architecture for anOpen Sciencemolecular compound database

Egon Willighagen, @egonwillighagenDept. of Bioinformatics - BiGCaT - Maastricht University

orcid.org/0000-0001-7542-0286

ACS New Orleans, 9 April 2013, #ACSNola

Department of Bioinformatics - BiGCaT 2

This session: Public Databases ...

• Public: what's that?– free access?– redistribute?–Modify?

• BTW, what is “Open Access” ???

Department of Bioinformatics - BiGCaT 3

This session: Serving the community...

• Service–What do people want?–Do they know what is possible?

• Community–Who are they? Personas!→–Usability must include learnability

Department of Bioinformatics - BiGCaT 4

Personas

• Not every scientist is alike• You cannot and must not target one

user• Instead, target at least 2 different

users, particularly:–The hacker doing all the actual

bioinformatics in the lab–The professor who has too little time to

understand things outside his narrow field

Department of Bioinformatics - BiGCaT 5

Reason #1: Bioclipse decision support

Spjuth, O. et al. JCIM 2011 51(8):1840-1847.

Department of Bioinformatics - BiGCaT 6

Data #1: Linked Open Drug Data

M. Samwald, et al, Linked open drug data for pharmaceutical research and development, 2011, JChemInf.

Department of Bioinformatics - BiGCaT 7

Data^2: Linked Open Data

Linking Open Data cloud diagram, by Richard Cyganiak and Anja Jentzsch. http://lod-cloud.net/ Sept 2011, CC-BY-SA.

Department of Bioinformatics - BiGCaT 8

Linked Open Data in the Life Sciences

Department of Bioinformatics - BiGCaT 9

WikiPathways

Pico, AR et al. PLoS biology 6.7 (2008): e184.

Department of Bioinformatics - BiGCaT 10

PathVisio: Pathway Analysis

Van Iersel, M. et al. BMC Bioinfo. 2008 9(1):399.

Department of Bioinformatics - BiGCaT 11

Reason #2: Publishing

• Journals will increasingly require data deposition–e.g. BioMed Central:

Department of Bioinformatics - BiGCaT 12

Needs

• We must propagate rights–whether open or not!

• We must make things explicit–e.g. by using semantics–e.g. by using the InChI

Department of Bioinformatics - BiGCaT 13

Tool #1: licensing

Department of Bioinformatics - BiGCaT 14

Open Data #1: crystallography

Department of Bioinformatics - BiGCaT 15

Open Data #2: Open Notebook Science

Department of Bioinformatics - BiGCaT 16

Open Data #3: CrystalEye

Department of Bioinformatics - BiGCaT 17

Licensing Open not Required→

• But not providing info is a killer–no, not really because

no scientist seems to care

–yes, because how will a machine do? Think scalability and massive data integration efforts

Department of Bioinformatics - BiGCaT 18

Why does explicit licensing matter?

Because when there is a fire, you want immediate access to the fire hose. You do not want to wait for permission from the mayor.

Because when you like to validate your scientific results, you want immediate access to related data. You do not want to wait for permission from that professor who is on a conference tour for the next 4 weeks. You must have an immediate answer, whatever it is.

Department of Bioinformatics - BiGCaT 19

Tool #2: Semantic Web to the rescue

• Allows provenance–provide where data came from– tells us our rights

Department of Bioinformatics - BiGCaT 20

App #1: Spidering the semantic web

Spjuth, O et al. JChemInf 2013 5:14.

Department of Bioinformatics - BiGCaT 21

App #2: Making a web

http://rdf.openmolecules.net/

Department of Bioinformatics - BiGCaT 22

App #3: Open PHACTS Explorer

http://www.openphacts.org/ → room 349, 2:20pm

Department of Bioinformatics - BiGCaT 23

How #1: RDF Graphs

Department of Bioinformatics - BiGCaT 24

How #1: RDF Graphs

PREFIX cheminf: <http://semanticscience.org/resource/>

SELECT ?graph ?p ?o WHERE { GRAPH ?graph { ?mol cheminf:CHEMINF_000200 [ a cheminf:CHEMINF_000059 ; cheminf:SIO_000300 "$inchikey" ] ; ?p ?o . }}

Department of Bioinformatics - BiGCaT 25

NanoPub.org

Department of Bioinformatics - BiGCaT 26

Graph output

orcid.org/0000-0001-7542-0286

Department of Bioinformatics - BiGCaT 27

Is that it?!? Just an architecture??

Yes, but a simple and flexible one. Keep an eye out on my blog. This will happen in the next few months:

1. Aggregate all CCZero/PDDL data around chemical properties

1.Open Notebook Science (solubility, melting point)

2.ChemPedia

3.Crystallography (COD, CrystalEye)

4....

2. Calculate molecular properties with the CDK (and release as CCZero)

3. Host on http://linkedchemistry.info/chembox

Department of Bioinformatics - BiGCaT 28

CHEMINF ontology

orcid.org/0000-0001-7542-0286

Hastings, J. et al. PLoS ONE 2011 6(10):e25513.

Department of Bioinformatics - BiGCaT 29

Architecture

Triple Store(e.g. Virtuoso)

Web server(HTML / RDF)

• Graphs• Explicit license

info• InChI/FixedH

Department of Bioinformatics - BiGCaT 30

/FixedH ?!?!

Department of Bioinformatics - BiGCaT 31

Conclusions & Outlook

• We must propagate rights–whether open or not!

• We must make things explicit–e.g. by using semantics–e.g. by using the InChI with FixedH

Department of Bioinformatics - BiGCaT 32

More information

• @egonwillighagen• http://chem-bla-ics.blogspot.com/• http://egonw.github.com/

• http://orcid.org/0000-0001-7542-0286