Introduction

[Cord Wiljes]

Dear CHEMINF-Group,

my name is Cord Wiljes. I just joined this group and would like to take the opportunity to introduce myself: I studied chemistry at Bielefeld University (Germany) some years ago. Since then I worked for several years as an IT consultant. A few months ago I returned to the university and am working now at CITEC on a PhD thesis. CITEC is a special research project, in which about 100 scientists from various faculties are cooperating on the topic of human-machine interaction.

My goal is to investigate and develop methods to publish research data from natural sciences as linked data. Currently I am looking into chemical ontologies, so I am very happy I have found CHEMINF and this group! I would be very interested in discussions about chemistry on the Semantic Web, developing chemical ontologies and publishing linked chemical data.

To start with I have a question: Are there possibly any linked data code examples which use the CHEMINF terminology? Such sample code would be very useful to me to get a deeper understanding of CHEMINF and how to apply it.

Best regards,
Cord

-- 
Cord Wiljes
Semantic Computing Group
Excellence Cluster - Cognitive Interaction Technology (CITEC)
Bielefeld University
www.sc.cit-ec.uni-bielefeld.de

[Janna Hastings]

Dear Cord,

welcome to the group! Great to hear from you. For examples, you might
want to see our paper on the topic:
www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0025513.
One important distinction is that the CHEMINF ontology is for chemical
*information* entities such as descriptors and other metadata.
Chemical entities are described in the ChEBI ontology
(http://www.ebi.ac.uk/chebi).

Various other projects are currently starting to use CHEMINF for
standardisation of their descriptors in the linked data space. Egon
Willighagen (also copied in the discussion group) may be able to tell
you more...

Best wishes
Janna

[Cord Wiljes]

Dear Janna,

thank you for the explanation and the link to the paper!

I am still struggling to understand the meaning and correct usage of the term "chemical information entity". To give an example: I want to describe the boiling point of the chemical substance "water" .

For water I can use a ChEBI-URI: http://bio2rdf.org/chebi:15377
And the "boiling point descriptor" of CHEMINF: http://semanticscience.org/resource/CHEMINF_000257

Would the following statements be correct (for better readability I used the CHEMINF-labels instead of the URIs):

chebi:15377   :has-attribute   _:id1 .
_:id1   rdf:type   :boiling-point-descriptor .
_:id1   :has-value   "373.15 Kelvin" .

A "boiling point descriptor" is a chemical information entity. So - if I understand the term "information entity" correctly - it refers not to the actual (real-life) boiling point of a substance but rather to the boiling point's (virtual) representation in a model (or a computer). So while a substance can have only one boiling point, it can have multiple boiling point descriptors (?) But then I don't understand why the boiling point descriptor can be connected directly to the substance instead of being connected to the substance's boiling point.

And an additional question: Is there a way to make the semantics of the descriptor's unit "Kelvin" explicit? Something like

_:id1   :has-value _:id2 . 
_:id2   ex:hasNumericalValue   "373.15"^^xsd:float .
_:id2   ex:hasUnit   ex:Kelvin .

Please excuse if these questions may have been discussed before or might seem trivial or nitpicky. I am still quite at the beginning of understanding ontologies and the semantic web. So any help, hints or pointers would be greatly appreciated.

Best wishes,
Cord

[Egon Willighagen]

On Fri, Feb 17, 2012 at 9:02 PM, Janna Hastings
<janna.h...@gmail.com> wrote:
> Various other projects are currently starting to use CHEMINF for
> standardisation of their descriptors in the linked data space. Egon
> Willighagen (also copied in the discussion group) may be able to tell
> you more...

I am aggregating examples on this wiki page:

http://code.google.com/p/semanticchemistry/wiki/Examples

Some projects where I am (starting to) use CHEMINF:

- ChEMBL-RDF
- Blue Obelisk Descriptor Ontology (alignment; followed by CDK, Bioclipse)

Egon

--
Dr E.L. Willighagen
Postdoctoral Researcher
Department of Bioinformatics - BiGCaT
Maastricht University (http://www.bigcat.unimaas.nl/)
Homepage: http://egonw.github.com/
LinkedIn: http://se.linkedin.com/in/egonw
Blog: http://chem-bla-ics.blogspot.com/
PubList: http://www.citeulike.org/user/egonw/tag/papers

[Cord Wiljes]

P.S.: I see that there is a fifference between "chemical quality" a "chemical descriptor". For example there is a "refractive index" (CHEMINF_000228) which is a "chemical quality" and a "refractive index descriptor" (CHEMINF_000253) which is a "chemical descriptor".

Can a "chemical descriptor", being an artifact, have a dc:creator, while a "chemical quality" can not?

What is the rdfs:domain of the property "has value" (CHEMINF_000012)? Is it "chemical quality" or "chemical descriptor"?

Best wishes,
Cord

-- 
Cord Wiljes
Semantic Computing Group
Excellence Cluster - Cognitive Interaction Technology (CITEC)
Bielefeld University

Phone: +49 521 106 12036
Mail:  cwi...@cit-ec.uni-bielefeld.de
Web:   http://www.sc.cit-ec.uni-bielefeld.de
Room H-123
Morgenbreede 39
33615 Bielefeld

[Egon Willighagen]

On Tue, Feb 21, 2012 at 10:59 AM, Cord Wiljes
<cwi...@cit-ec.uni-bielefeld.de> wrote:
> P.S.: I see that there is a fifference between "chemical quality" a
> "chemical descriptor". For example there is a "refractive index"
> (CHEMINF_000228) which is a "chemical quality" and a "refractive index
> descriptor" (CHEMINF_000253) which is a "chemical descriptor".
>
> Can a "chemical descriptor", being an artifact, have a dc:creator, while a
> "chemical quality" can not?

The first is something inherent to the entity (and can be 'measured'),
the second is calculated.

[Cord Wiljes]

Am 21.02.2012 11:01, schrieb Egon Willighagen:
> On Tue, Feb 21, 2012 at 10:59 AM, Cord Wiljes
> <cwi...@cit-ec.uni-bielefeld.de> wrote:
>> P.S.: I see that there is a fifference between "chemical quality" a
>> "chemical descriptor". For example there is a "refractive index"
>> (CHEMINF_000228) which is a "chemical quality" and a "refractive index
>> descriptor" (CHEMINF_000253) which is a "chemical descriptor".
>>
>> Can a "chemical descriptor", being an artifact, have a dc:creator, while a
>> "chemical quality" can not?
> The first is something inherent to the entity (and can be 'measured'),
> the second is calculated.
>
> Egon

Calculated from molecular structure data? So using the "boiling point
descriptor" would imply that the the object of "has value" is not
measured but calculated?

Best wishes,
Cord

[Egon Willighagen]

On Tue, Feb 21, 2012 at 11:51 AM, Cord Wiljes
<cwi...@cit-ec.uni-bielefeld.de> wrote:
> Calculated from molecular structure data? So using the "boiling point
> descriptor" would imply that the the object of "has value" is not measured
> but calculated?

Yes...

Checking that Examples wiki page, it seems there are two 'has value' predicates:

SIO_000300 has value
CHEMINF_000012 has value

The first does not seem to have domain or range defined:
http://semanticscience.org/resource/SIO_000300

The CHEMINF one I do not fully understand at first sight:
http://semanticscience.org/resource/CHEMINF_000012

Janna, Michel ?

[Cord Wiljes]

Dear Egon,

thank you for providing these code examples, which are very helpful. I noticed that all examples are referring to molecules. CHEMINF contains chemical descriptors which are referring to substances (like "boiling point descriptor" CHEMINF_000257) while others are referring to molecules (like "DL stereochemical descriptor" CHEMINF_000051) Is there a way to differentiate between these two categories? And could you imagine a way to connect them? Something like a property pair isMoleculeToSubstance / isSubstanceToMolecule

:water  :isSubstanceToMolecule   :waterMolecule .
:water  a  :substance .
:waterMolecule  a  :molecule .

Best wishes,
Cord

[Egon Willighagen]

On Tue, Feb 21, 2012 at 12:23 PM, Cord Wiljes
<cwi...@cit-ec.uni-bielefeld.de> wrote:
> thank you for providing these code examples, which are very helpful. I
> noticed that all examples are referring to molecules. CHEMINF contains
> chemical descriptors which are referring to substances (like "boiling point
> descriptor" CHEMINF_000257) while others are referring to molecules (like
> "DL stereochemical descriptor" CHEMINF_000051) Is there a way to
> differentiate between these two categories?

Yes, a molecule is a single chemical structure, like a single benzene
molecule, as you describe below. Many (benzene) molecules form a
substance, a pure substance if consisting of one kind of molecule, a
mixed substance if consisting of two or more different molecules. A
boiling point is not a property of a single molecule, but the
temperature at which a bunch of molecules (a substance) goes from a
fluid phase to a gas phase.

> And could you imagine a way to
> connect them? Something like a property pair isMoleculeToSubstance /
> isSubstanceToMolecule
>
> :water  :isSubstanceToMolecule   :waterMolecule .
> :water  a  :substance .
> :waterMolecule  a  :molecule .

"part of" comes to mind... but I am not sure if we have a predicate
for that yet...

Janna, Michel, Nico, do we?

[Frank Gibson]

Comments in line

A molecule of a substance is slightly more specific than part_of - if you remove a molecule you have not destroyed the water_substance. The relationship here you may be looking for may be _grain_ and then the notions of portions_of_substance which are a collection of multiple instances of _grain_

So that

Water_substance has_grain =>1 some H20_molecule

More detail on this topic can be found here

http://www.cs.man.ac.uk/~rector/papers/Granularity-scale-collectivity-rector.pdf

Cheers

Frank

Egon

--
Dr E.L. Willighagen
Postdoctoral Researcher
Department of Bioinformatics - BiGCaT
Maastricht University (http://www.bigcat.unimaas.nl/)
Homepage: http://egonw.github.com/
LinkedIn: http://se.linkedin.com/in/egonw
Blog: http://chem-bla-ics.blogspot.com/
PubList: http://www.citeulike.org/user/egonw/tag/papers

--
Frank Gibson, PhD
http://www.fgibson.com

[Cord Wiljes]

Hi Frank,

thanks for pointing out this intersting paper! So

water_molecule   has_part   hydrogen_atom
water_substance   has_grain  water_molecule
water_molecule     has_property  [a  molecular_weight]
water_substance   has_property  [a  freezing_point]

sounds fine to me.

Best wishes,
Cord

[Michel Dumontier]

Hi Egon, Cord,

  CHEMINF's 'has value' property followed the pattern set out for SIO - that there is only one datatype property which captures the actual quantity. In specifying CHEMINF-based descriptors, we recommend the use of CHEMINF's 'has value', and that this would map as a subproperty of SIO's.

m.

--

Michel Dumontier

Visiting Associate Professor, Stanford University

Associate Professor of Bioinformatics, Carleton University
http://dumontierlab.com

[Michel Dumontier]

Hi Cord,

  SIO provides predicates to describe finer mereological relationships, like that between a substance composed of certain molecules.

 'has direct part' ( http://semanticscience.org/resource/SIO_000273 ) is meant to indicate a part at a particular level of granularity - it is OWL-DL safe, in that it is non-transitive so you can use the relation in a cardinality restriction.

'has component part' (  http://semanticscience.org/resource/SIO_000369 ) is meant to indicate that a part is necessary to the identity of the whole, such that if you remove the part, the whole loses it's identity. So this is an to describe the relation between a molecule and a component atom.

Best,

m.