Re: HCLS knowledgebase extension: PDSP Ki database uploaded to DERI triplestore

This is wonderful, Matthias.

We need to look at how this jibes with what we've created in BIRN &  
NIF for the IUPHAR classification for G-protein coupled receptors.   
We definitely want our representations to stay commensurate.  We've  
got this in the ontology driving integration between SenseLab, CCDB,  
and NeuroMorpho.org.  This includes the complete set of NeuroName  
brain regions refactored for a BFO+OBO-RO + PATO representation in  
OWL (BIRNLex-Anatomy), nerve cell types (NIF-Cell.owl) which we are  
working to make compatible with the OBO CL ontology, and OBI.

For the IUPHAR representation, I'm using the Sequence ontology, BFO,  
OBO-RO, and PATO.  I also use NCBI eUtils to draw out multiple  
organism representations using Homologene, as well as to assemble all  
the appropriate parts from protein superfamily to the organism- 
specific protein (think UniProt) on through to transcript, gene, and  
chromosome, so that you could resolve a SPARQL query that asks:

	What Cav2-type Ca++-Channels are on mouse chromosome 11?

I've got ligands in there - at least those the IUPHAR group has  
listed for the receptor families.  It's nothing like the variety in  
PSPD Ki (7500 test ligands).  There are some blank node problems with  
my current representation of the ligand interactions that I have to  
fix, but it does use the BFO dependent continuants to describe a  
"disposition to bind" a ligand which is then linked to specific  
ligands (e.g., the many Adenosine analogs that bind to the various  
Adenosine receptors).  I'm not certain whether disposition is the way  
to do this - or whether there should be a "ligand role" that inheres  
in the ligand molecules.  You're still left trying to figure out how  
to link the ligands to the receptors.  The alternative is to  
represent the binding process - which you are doing very nicely here  
using the generic GO "binding" molecular function.  Since IUPHAR is  
receptor-centric, representing binding as a disposition of the  
receptors, seemed reasonable as a first pass on IUPHAR.

The other issue I've still not worked out is in order to be able to  
infer the set described in the example above, all the classes need to  
have the proper disjoint relations and closing axioms.  I can easily  
do this, since this whole representation is algorithmically generated  
off an algorithmic dump of the IUPHAR spreadsheets using Jena & the  
NCBI eUtils.  The problem was, when I put the sibling disjoints in  
there for the mRNAs & genes (essentially 500+ sibling descendants of  
SO:mRNA and SO:multi-cistronic gene), I couldn't open the file in  
Protege v3.3.1, so I had to drop those for the time being.

One comment on the PSPD Ki representation you have:
	Do you have the tissue & organism source fields from PSPD Ki?  If  
these could be included in the result list (maybe even used to sort  
the list - in addition to sorting on ligand), then it would be clear  
why there are multiple entries for some of the ligands.

Again - very nice work.

Cheers,
Bill

P.S.: We've also run the same conversion on the IUPHAR voltage-gated  
ion channels.  Both of these will be reviewed at the upcoming Protege  
Ontology (PRO) meeting to figure out how to make use of Sequence  
Ontology & the PRO components (ProEvo & ProForm) - as well as the RNA  
Ontology.

As I mentioned about 2 months ago, these files are available for  
review at:

	BIRNLex ontology (some components re-used in NIF ontology):

		http://purl.org/nbirn/birnlex/ontology/birnlex.owl

	NIF ontology:
	
		http://purl.org/nif/ontology/nif.owl

On Oct 31, 2007, at 11:10 AM, samwald@gmx.at wrote:

>
> I have uploaded the 'beta version' of the OWL conversion of the  
> PDSP Ki database to the SPARQL endpoint hosted by DERI.
>
> The address is:
> http://hcls.deri.ie/sparql
>
> Alternatively, you can also use the SPARQL endpoint of the Yale  
> Center for Medical Informatics (only contains the SenseLab  
> ontologies):
> http://neuroweb.med.yale.edu:8890/sparql
>
> Since the datasets makes use of the SenseLab URIs for receptors,  
> the queries can also be connected to further information about  
> neuronal cell types, cellular properties, brain regions etc.
>
> A possible SPARQL query showing some ligands of the serotonin  
> receptor 5-HT2A, together with their affinity (lower values mean  
> higher affinity, i.e. the substance is of higher pharmacological  
> interest):
>
> =====
>
> PREFIX neurondb: <http://purl.org/ycmi/senselab/neuron_ontology.owl#>
> PREFIX obo_essentials: <http://purl.org/zen/obo_essentials.owl#>
> PREFIX ro:
> <http://www.obofoundry.org/ro/ro.owl#>
> PREFIX rdfs: <http://www.w3.org/2000/01/rdf-schema#>
>
> SELECT ?name_of_ligand ?ki_value FROM <http://purl.org/ycmi/ki/ 
> core.owl> WHERE  {?nicotinic_receptor a neurondb:_5-HT2A .
> ?ligand a ?ligand_class .
> ?process a obo_essentials:GO_0005488_process .
> ?process ro:has_participant ?nicotinic_receptor .
> ?process ro:has_participant ?ligand .
> ?ligand_class rdfs:label ?name_of_ligand .
> ?ligand ro:bearer_of ?quality .
> ?quality rdf:value ?ki_value .
>  }
>
> =====
> First three results:
>
> name_of_ligand -- ki_value
> --------------------------
> tryptamine -- 218.7761623949
> (R)-noradrenaline --455970.15
> yohimbine -- 4790
> chlorpromazine -- 1.8
>
> Some of the ligands have the name 'null', this still needs to be  
> fixed.
>
> The PDSP Ki datasets can be mirrored on other installations of  
> Virtuoso with the following iSQL command:
>
> DB.DBA.RDF_LOAD_RDFXML(
> xml_uri_get('http://neuroweb.med.yale.edu/svn/trunk/ontology/ki/ 
> core.owl', 'http://neuroweb.med.yale.edu/svn/trunk/ontology/ki/ 
> core.owl'),
> 'http://purl.org/ycmi/ki/core.owl', 'http://purl.org/ycmi/ki/core.owl'
> )
>
>
> Please note that the current HCLS demo knowledge base contains  
> outdated namespaces for the SenseLab ontologies (we are using PURLs  
> now), so the PDSP Ki datasets do not readily connect to the rest of  
> the knowledge base. The namespaces in the demo KB will soon be  
> updated (at least I would like to do so).
>
> cheers,
> Matthias Samwald
> -- 
> Der GMX SmartSurfer hilft bis zu 70% Ihrer Onlinekosten zu sparen!
> Ideal für Modem und ISDN: http://www.gmx.net/de/go/smartsurfer
>



William Bug, M.S., M.Phil.                                         		 
email: wbug@ncmir.ucsd.edu
Ontological Engineer/Programmer Analyst III		work: (858) 822-0739
Biomedical Informatics Research Network
Dept. of Neuroscience, School of Medicine
University of California, San Diego
9500 Gilman Drive
La Jolla, CA 92093

Please note my email has recently changed