Here is a small excerpt of the outcome of my SIDER conversion:
http://hcls.deri.org/atag/data/SIDER_atags_excerpt.html
Of course, listing aTags on a page in this manner is not very
human-friendly, they should rather be embedded in a website with better
navigation, split into several web pages. Note that you could just
copy&paste some of the aTags to another HTML area, mix it with other RDFa,
and the data would be carried with it ('client-side mashup'). I also encoded
basic copyright and authorship information in RDFa.
Triples in this page, extracted with Triplr.org:
http://triplr.org/html/hcls.deri.org/atag/data/SIDER_atags_excerpt.html
(or shorter: http://tinyurl.com/cerwpp)
This is just a small excerpt, the entire file is several megabytes large. I
still need to fix some issues with XML Literals in RDFa, and will then add
the resulting triples of the full document to the HCLS KB.
Cheers,
Matthias
--------------------------------------------------
From: "Peter Ansell" <ansell.pe...@gmail.com>
Sent: Wednesday, April 29, 2009 11:26 AM
To: "Matthias Samwald" <samw...@gmx.at>
Cc: "public-semweb-lifesci" <public-semweb-lifesci@w3.org>
Subject: Re: drug side effects
2009/4/29 Matthias Samwald <samw...@gmx.at>:
I have also started to convert SIDER into aTags, but our two conversions
will be complementary. What I am doing is mapping the information to the
OBO
disease ontology and to DBpedia (for drug names), and reusing OBO and
DBpedia URIs directly. The conversion will be lossy when no matches are
found in these resources, but it will be bound more tightly to OBO and
DBpedia.
I will have to investigate how aTags works more deeply. I developed my
own tagging vocabulary [1] but it hasn't seen many uses so far. From a
rough view, the terms:hasEvidence URI seems to map well to aTags
although it might be a bit specialised. Is [2] the only place aTags
are described?
The STITCH ids for chemicals seem to be equivalent to Pubchem in all
cases I
checked. Are there deviations?
The STITCH and SIDER notes seem to indicate that they have reused the
pubchem identifiers, so owl:sameAs should be appropriate to
http://bio2rdf.org/pubchem:STITCHID.
By the way, some time back I also found two other datasets about drug
side
effects / pharmacovigilance data that could be used for conversions.
Maximum Recommended Therapeutic Dose (MRTD) Database
http://www.fda.gov/Cder/Offices/OPS_IO/MRTD.htm
Adverse Event Reporting System (AERS) Statistics
http://www.fda.gov/cder/aers/extract.htm
I will try to analyze those to see if they are worthwhile candidates for
conversion. The AERS data is a bit complicated, and converting only some
of
the data might suffice.
Partial rdfisations are still useful. If you release the script you
used to do it someone else can improve the coverage in the future if
needed.
Adverse events seem to be generally identified with terms from the MedDra
dictionary (http://www.meddramsso.com), which is proprietary. I find this
a
bit irritating -- the entry barriers for institutions to create reports
about adverse events should be as low as possible, which is not really
helped by locking away the standard dictionary for codifying adverse
events...
It would be interesting to see what relationships are possible between
drugbank and these datasets, as drugbank is linked directly into a
large number of Bio2RDF datasets so the bridge from that point is
made.
Cheers,
Peter
[1] http://bio2rdf.mquter.qut.edu.au/ns/terms
[2] http://esw.w3.org/topic/HCLSIG_BioRDF_Subgroup/aTags
