sbml-annot — SBML L3 annotations discussion list

[sbml-annot] SBML annot

[Neil S. <nei...@ma...>]

Hi all,

You may remember some time back in the early 16th Century, we were working on a proposal for a new SBML annot package.

I apologise for not giving this my undivided attention over the last couple of months, but have made a couple of updates to some of the pages in an attempt to provoke further discussion that will hopefully ultimately bang these docs into shape.

http://sbml.org/Events/Other_Events/Annotation_package_workshop_2010/Draft_Proposal_for_the_Annot_Package

Updates are highlighted with virtual fluorescent yellow marker pen.

Cheers,

Neil.

Neil Swainston
Experimental Officer

Manchester Centre for Integrative Systems Biology
Manchester Interdisciplinary Biocentre
University of Manchester
Manchester M1 7DN
England

Re: [sbml-annot] negation in SBML annotations

[Nicolas Le n. <le...@eb...>]

On 21/06/10 19:06, Neil Swainston wrote:

> Another thing that came up was my approach of putting all of this into
> rdf. I didn't really do this conciously, but it seems to me a good idea.
> Why is a hybrid rdf / non-rdf solution preferable?

There are two aspects: rdf / non-rdf
but also SBML / generic

The current SBML controlled annotation scheme (L3 core) does not actually 
contains SBML elements or attributes. As such, it is considered for 
adoption by other languages, such as CellML or NeuroML. IMHO, we should 
provide with the annot package a cleaner, more versatile version, that 
would still be independent of SBML. It will not be just RDF, if only 
because of the semantics attached to the bags. But there won't be elements 
or attributes that are SBML-specific.

>
> Cheers,
>
> Neil.
>
>
> Neil Swainston
> Experimental Officer
>
> Manchester Centre for Integrative Systems Biology
> Manchester Interdisciplinary Biocentre
> University of Manchester
> Manchester M1 7DN
> England
>
> On 21 Jun 2010, at 13:06, Dagmar Waltemath
> <dag...@un...> wrote:
>
>> Hej Sarah (and all),
>>
>> thanks a lot for that nice summary :-)
>> Let me just add some points.
>>
>> Sarah Keating wrote:
>>> Neil is right, the thread is getting hard to read BUT my experience
>>> of Google Wave is that it becomes equally disjointed.
>>>
>>> This is a summary of my understanding of where we are at:
>>>
>>> 1)
>>> We need to represent two negatives
>>> a) the concious absence of something from the model
>>> b) the fact that something in the model is NOT another thing
>>>
>>> Everyone agrees on this.
>>>
>>> 2)
>>> The current proposed solution for a) is along the lines of
>>>
>>> <listOfSpecies>
>>> <annotation>
>>> <annot:listOfAbsentEntitities>
>>> <annot:ommittedEntity metaid="foo"/>
>>> <annot:ommittedEntity metaid="bar" />
>>> </annot:listOfAbsentEntitites>
>>> <rdf:RDF>
>>> <rdf:description rdf:about="foo">
>>> <bqbiol:is>
>>> <rdf:Bag>
>>> <rdf:li rdf:resource="urn:miriam:obo.go:GO%3A11111111"/>
>>> </rdf:Bag>
>>> </bqbiol:is>
>>> </rdf:description>
>>> <rdf:description rdf:about="bar">
>>> [..]
>>> </rdf:description>
>>> </annotation>
>>> </rdf:RDF>
>>> </listOfSpecies>
>>>
>>>
>>> I think most people seem happy with this but there is still debate on
>>> whether this should be placed on the Model element or the individual
>>> ListOfXYZ elements.
>>>
>>
>> Regarding this debate, I am afraid I cannot imagine how we will know
>> *what* the omitted entity actually *is* if we put the
>> listOfAbsentEntitites on the model. I have looked at Neils example,
>> but the only way to figure out would be to resolve the URN and then
>> guess, would it not? There was no connection to particular SBML
>> constructs from the list of absent objects, was there?
>> I do think we need to identify the SBML concept for each omitted
>> entity. The best example I could think of was a GO entry that is
>> allowed to represent a species in SBML, but is not allowed to
>> represent a parameter. I thought that it might become hard to reason
>> about what the omitted species refers to if we do not specify that in
>> the model.
>>
>> One point that was raised in line with the omitted entities was that
>> we might want to restrict the qualifiers that are allowed. Otherwise
>> it might get very complicated (the omittedEntity hasPart something)
>> ... that's the same concern as Sarah's further down for the negations.)
>> Maybe allowing a simple "is" will be sufficient?
>>
>> Another point was the mention of a possible (optional) compartment
>> attribute.
>>
>> <annot:listOfAbsentEntitities>
>> <annot:ommittedEntity metaid="foo" compartment="comp_metaid"/>
>> <annot:ommittedEntity metaid="bar" />
>> </annot:listOfAbsentEntitites>
>>
>>
>> I must say I did not get what the conclusion was for Christian's water
>> example (although I could see it was hard to implement what he wanted
>> to say). But I would be happy if someone could clarify and/or give
>> examples :-)
>>
>>> 3) There are two proposed solutions for b)
>>>
>>> Solution 1: We create negative identifiers such as 'isNot'
>>> 'doesNotHavePart' etc.
>>>
>>> Solution 2: We use a boolean attribute on the identifier to indicate
>>> whether we are negating the statement:
>>> <bqbiol:is negation="true"> represents isNOT
>>>
>>
>> Here I also (still) prefer solution 2. And I'll leave the thinking
>> about what the different negated qualifiers mean to Neil who might
>> have loads of time to do so after the 23rd :-P
>>
>> Cheers,
>> Dagmar
>>
>>
>>> ******************
>>> My thoughts:
>>>
>>> I think the construct of 2 is good but have no real preference on the
>>> model/listOf issue.
>>>
>>> On 3.
>>> Given that the issue of the semantics of the identifiers has already
>>> been raised I think Solution 1 is asking for more trouble. I like
>>> solution 2 but again this needs careful thought as to how it evolves
>>> to a statement "subject has NOT predicate whose value is object" :-)
>>> This may give Neil even more sleepless nights than the current
>>> performance of the England football team.
>>>
>>> Sarah


-- 
Nicolas LE NOVERE, Computational Neurobiology, EMBL-EBI, Wellcome-Trust
Genome Campus, Hinxton CB101SD UK, Mob:+447833147074, Tel:+441223494521
Fax:468,Skype:n.lenovere,AIM:nlenovere,MSN:nle...@ho...(NOT email)
http://www.ebi.ac.uk/~lenov/, http://www.ebi.ac.uk/compneur/, @lenovere

Re: [sbml-annot] negation in SBML annotations

[Stefan H. <sh...@vb...>]

Hello Neil,

On Mon, 21 Jun 2010 08:22:01 +0100
Neil Swainston <nei...@ma...> wrote:

> Hi Stefan,
> 
> The problem with enforcing annotation is that sometimes we either  
> simply don't know what the thing is, or that we don't but we can't  
> represent it in MIRIAM-flavoured RDF.
> 
> The latter we should fix in the new annot package, but the former  
> remains a problem.

Why? 
'xxx is unknown' is a perfectly valid triplet :)
It is in my eyes a valid annotation and we should support it. We may
need to add a general MIRIAM resource which expresses the concept of
unknown, as statements like: 'xxx has unknown' are also conceivable.

Thanks,
Stefan


-- 
Stefan Hoops, Ph.D.
Senior Project Associate
Virginia Bioinformatics Institute - 0477
Virginia Tech
Bioinformatics Facility II
Blacksburg, Va 24061, USA

Phone: (540) 231-1799
Fax:   (540) 231-2606
Email: sh...@vb...

Re: [sbml-annot] negation in SBML annotations

[Neil S. <nei...@ma...>]

Afternoon,

>The only thing RDF can express are triples. For the negative knowledge, we may want to attach all kinds of things to it. First will be annotations and notes.

Why is this a problem specific to negation? We agreed to chain triples together for non-negative (positive!) knowledge. Why is the negative case any different? 

listOfSpecies has omittedSpecies X
X has .... whatever

This seems to be OK to me. Am I missing something?

Cheers,

Neil.

Neil Swainston
Experimental Officer

Manchester Centre for Integrative Systems Biology
Manchester Interdisciplinary Biocentre
University of Manchester
Manchester M1 7DN
England

On 24 Jun 2010, at 09:08, Nicolas Le novère wrote:

> On 21/06/10 19:06, Neil Swainston wrote:
> 
>> Another thing that came up was my approach of putting all of this into
>> rdf. I didn't really do this conciously, but it seems to me a good idea.
>> Why is a hybrid rdf / non-rdf solution preferable?
> 
> The only thing RDF can express are triples. For the negative knowledge, we 
> may want to attach all kinds of things to it. First will be annotations and 
> notes.
> 
> Actually, the more I dig into the package, the more I realise the 
> limitations of RDF. And I think we are about certain that we won't be able 
> to use "pure" RDF, without added semantics coming from the package 
> specification or additional attributes.
> 
> I had a meeting recently with the developers of VPH's project RICORDO, and 
> they were puzzled by the usage of Bags. And after thinking more (and I am 
> conscious that Allyson realised that long ago), it is true that the 
> semantics carried by a bag is not part of RDF, but of SBML annotations:
> 
> bqbiol:part
>   Bag
>     X
>     Y
> 
> Really means *in RDF*
> 
> bqbiol:part X
> bqbiol:part Y
> 
> But I believe we should add an SBML annot layer of semantics to the RDF, in 
> order to relate the triples together. For instance, to introduce 
> "disjointPart", for solving one of the expressivity problems discussed at 
> the meeting,
> 
> bqbiol:disjointPart
>   Bag
>     X
>     Y
> 
> The meaning of the "disjoint" covers the whole bag. Actually, an 
> alternative is the approach supported by Dagmar and Sarah for negation:
> 
> bqbiol:part
>   Bag disjoint="true"
>     X
>     Y
> 
> 
>> Neil Swainston
>> Experimental Officer
>> 
>> Manchester Centre for Integrative Systems Biology
>> Manchester Interdisciplinary Biocentre
>> University of Manchester
>> Manchester M1 7DN
>> England
>> 
>> On 21 Jun 2010, at 13:06, Dagmar Waltemath
>> <dag...@un...> wrote:
>> 
>>> Hej Sarah (and all),
>>> 
>>> thanks a lot for that nice summary :-)
>>> Let me just add some points.
>>> 
>>> Sarah Keating wrote:
>>>> Neil is right, the thread is getting hard to read BUT my experience
>>>> of Google Wave is that it becomes equally disjointed.
>>>> 
>>>> This is a summary of my understanding of where we are at:
>>>> 
>>>> 1)
>>>> We need to represent two negatives
>>>> a) the concious absence of something from the model
>>>> b) the fact that something in the model is NOT another thing
>>>> 
>>>> Everyone agrees on this.
>>>> 
>>>> 2)
>>>> The current proposed solution for a) is along the lines of
>>>> 
>>>> <listOfSpecies>
>>>> <annotation>
>>>> <annot:listOfAbsentEntitities>
>>>> <annot:ommittedEntity metaid="foo"/>
>>>> <annot:ommittedEntity metaid="bar" />
>>>> </annot:listOfAbsentEntitites>
>>>> <rdf:RDF>
>>>> <rdf:description rdf:about="foo">
>>>> <bqbiol:is>
>>>> <rdf:Bag>
>>>> <rdf:li rdf:resource="urn:miriam:obo.go:GO%3A11111111"/>
>>>> </rdf:Bag>
>>>> </bqbiol:is>
>>>> </rdf:description>
>>>> <rdf:description rdf:about="bar">
>>>> [..]
>>>> </rdf:description>
>>>> </annotation>
>>>> </rdf:RDF>
>>>> </listOfSpecies>
>>>> 
>>>> 
>>>> I think most people seem happy with this but there is still debate on
>>>> whether this should be placed on the Model element or the individual
>>>> ListOfXYZ elements.
>>>> 
>>> 
>>> Regarding this debate, I am afraid I cannot imagine how we will know
>>> *what* the omitted entity actually *is* if we put the
>>> listOfAbsentEntitites on the model. I have looked at Neils example,
>>> but the only way to figure out would be to resolve the URN and then
>>> guess, would it not? There was no connection to particular SBML
>>> constructs from the list of absent objects, was there?
>>> I do think we need to identify the SBML concept for each omitted
>>> entity. The best example I could think of was a GO entry that is
>>> allowed to represent a species in SBML, but is not allowed to
>>> represent a parameter. I thought that it might become hard to reason
>>> about what the omitted species refers to if we do not specify that in
>>> the model.
>>> 
>>> One point that was raised in line with the omitted entities was that
>>> we might want to restrict the qualifiers that are allowed. Otherwise
>>> it might get very complicated (the omittedEntity hasPart something)
>>> ... that's the same concern as Sarah's further down for the negations.)
>>> Maybe allowing a simple "is" will be sufficient?
>>> 
>>> Another point was the mention of a possible (optional) compartment
>>> attribute.
>>> 
>>> <annot:listOfAbsentEntitities>
>>> <annot:ommittedEntity metaid="foo" compartment="comp_metaid"/>
>>> <annot:ommittedEntity metaid="bar" />
>>> </annot:listOfAbsentEntitites>
>>> 
>>> 
>>> I must say I did not get what the conclusion was for Christian's water
>>> example (although I could see it was hard to implement what he wanted
>>> to say). But I would be happy if someone could clarify and/or give
>>> examples :-)
>>> 
>>>> 3) There are two proposed solutions for b)
>>>> 
>>>> Solution 1: We create negative identifiers such as 'isNot'
>>>> 'doesNotHavePart' etc.
>>>> 
>>>> Solution 2: We use a boolean attribute on the identifier to indicate
>>>> whether we are negating the statement:
>>>> <bqbiol:is negation="true"> represents isNOT
>>>> 
>>> 
>>> Here I also (still) prefer solution 2. And I'll leave the thinking
>>> about what the different negated qualifiers mean to Neil who might
>>> have loads of time to do so after the 23rd :-P
>>> 
>>> Cheers,
>>> Dagmar
>>> 
>>> 
>>>> ******************
>>>> My thoughts:
>>>> 
>>>> I think the construct of 2 is good but have no real preference on the
>>>> model/listOf issue.
>>>> 
>>>> On 3.
>>>> Given that the issue of the semantics of the identifiers has already
>>>> been raised I think Solution 1 is asking for more trouble. I like
>>>> solution 2 but again this needs careful thought as to how it evolves
>>>> to a statement "subject has NOT predicate whose value is object" :-)
>>>> This may give Neil even more sleepless nights than the current
>>>> performance of the England football team.
>>>> 
>>>> Sarah
> 
> 
> -- 
> Nicolas LE NOVERE, Computational Neurobiology, EMBL-EBI, Wellcome-Trust
> Genome Campus, Hinxton CB101SD UK, Mob:+447833147074, Tel:+441223494521
> Fax:468,Skype:n.lenovere,AIM:nlenovere,MSN:nle...@ho...(NOT email)
> http://www.ebi.ac.uk/~lenov/, http://www.ebi.ac.uk/compneur/, @lenovere
> 
> 
> 
> ------------------------------------------------------------------------------
> ThinkGeek and WIRED's GeekDad team up for the Ultimate 
> GeekDad Father's Day Giveaway. ONE MASSIVE PRIZE to the 
> lucky parental unit.  See the prize list and enter to win: 
> http://p.sf.net/sfu/thinkgeek-promo
> _______________________________________________
> Sbml-annot mailing list
> Sbm...@li...
> https://lists.sourceforge.net/lists/listinfo/sbml-annot

Re: [sbml-annot] negation in SBML annotations

[Nicolas Le n. <le...@eb...>]

On 21/06/10 19:06, Neil Swainston wrote:

> Another thing that came up was my approach of putting all of this into
> rdf. I didn't really do this conciously, but it seems to me a good idea.
> Why is a hybrid rdf / non-rdf solution preferable?

The only thing RDF can express are triples. For the negative knowledge, we 
may want to attach all kinds of things to it. First will be annotations and 
notes.

Actually, the more I dig into the package, the more I realise the 
limitations of RDF. And I think we are about certain that we won't be able 
to use "pure" RDF, without added semantics coming from the package 
specification or additional attributes.

I had a meeting recently with the developers of VPH's project RICORDO, and 
they were puzzled by the usage of Bags. And after thinking more (and I am 
conscious that Allyson realised that long ago), it is true that the 
semantics carried by a bag is not part of RDF, but of SBML annotations:

bqbiol:part
   Bag
     X
     Y

Really means *in RDF*

bqbiol:part X
bqbiol:part Y

But I believe we should add an SBML annot layer of semantics to the RDF, in 
order to relate the triples together. For instance, to introduce 
"disjointPart", for solving one of the expressivity problems discussed at 
the meeting,

bqbiol:disjointPart
   Bag
     X
     Y

The meaning of the "disjoint" covers the whole bag. Actually, an 
alternative is the approach supported by Dagmar and Sarah for negation:

bqbiol:part
   Bag disjoint="true"
     X
     Y


> Neil Swainston
> Experimental Officer
>
> Manchester Centre for Integrative Systems Biology
> Manchester Interdisciplinary Biocentre
> University of Manchester
> Manchester M1 7DN
> England
>
> On 21 Jun 2010, at 13:06, Dagmar Waltemath
> <dag...@un...> wrote:
>
>> Hej Sarah (and all),
>>
>> thanks a lot for that nice summary :-)
>> Let me just add some points.
>>
>> Sarah Keating wrote:
>>> Neil is right, the thread is getting hard to read BUT my experience
>>> of Google Wave is that it becomes equally disjointed.
>>>
>>> This is a summary of my understanding of where we are at:
>>>
>>> 1)
>>> We need to represent two negatives
>>> a) the concious absence of something from the model
>>> b) the fact that something in the model is NOT another thing
>>>
>>> Everyone agrees on this.
>>>
>>> 2)
>>> The current proposed solution for a) is along the lines of
>>>
>>> <listOfSpecies>
>>> <annotation>
>>> <annot:listOfAbsentEntitities>
>>> <annot:ommittedEntity metaid="foo"/>
>>> <annot:ommittedEntity metaid="bar" />
>>> </annot:listOfAbsentEntitites>
>>> <rdf:RDF>
>>> <rdf:description rdf:about="foo">
>>> <bqbiol:is>
>>> <rdf:Bag>
>>> <rdf:li rdf:resource="urn:miriam:obo.go:GO%3A11111111"/>
>>> </rdf:Bag>
>>> </bqbiol:is>
>>> </rdf:description>
>>> <rdf:description rdf:about="bar">
>>> [..]
>>> </rdf:description>
>>> </annotation>
>>> </rdf:RDF>
>>> </listOfSpecies>
>>>
>>>
>>> I think most people seem happy with this but there is still debate on
>>> whether this should be placed on the Model element or the individual
>>> ListOfXYZ elements.
>>>
>>
>> Regarding this debate, I am afraid I cannot imagine how we will know
>> *what* the omitted entity actually *is* if we put the
>> listOfAbsentEntitites on the model. I have looked at Neils example,
>> but the only way to figure out would be to resolve the URN and then
>> guess, would it not? There was no connection to particular SBML
>> constructs from the list of absent objects, was there?
>> I do think we need to identify the SBML concept for each omitted
>> entity. The best example I could think of was a GO entry that is
>> allowed to represent a species in SBML, but is not allowed to
>> represent a parameter. I thought that it might become hard to reason
>> about what the omitted species refers to if we do not specify that in
>> the model.
>>
>> One point that was raised in line with the omitted entities was that
>> we might want to restrict the qualifiers that are allowed. Otherwise
>> it might get very complicated (the omittedEntity hasPart something)
>> ... that's the same concern as Sarah's further down for the negations.)
>> Maybe allowing a simple "is" will be sufficient?
>>
>> Another point was the mention of a possible (optional) compartment
>> attribute.
>>
>> <annot:listOfAbsentEntitities>
>> <annot:ommittedEntity metaid="foo" compartment="comp_metaid"/>
>> <annot:ommittedEntity metaid="bar" />
>> </annot:listOfAbsentEntitites>
>>
>>
>> I must say I did not get what the conclusion was for Christian's water
>> example (although I could see it was hard to implement what he wanted
>> to say). But I would be happy if someone could clarify and/or give
>> examples :-)
>>
>>> 3) There are two proposed solutions for b)
>>>
>>> Solution 1: We create negative identifiers such as 'isNot'
>>> 'doesNotHavePart' etc.
>>>
>>> Solution 2: We use a boolean attribute on the identifier to indicate
>>> whether we are negating the statement:
>>> <bqbiol:is negation="true"> represents isNOT
>>>
>>
>> Here I also (still) prefer solution 2. And I'll leave the thinking
>> about what the different negated qualifiers mean to Neil who might
>> have loads of time to do so after the 23rd :-P
>>
>> Cheers,
>> Dagmar
>>
>>
>>> ******************
>>> My thoughts:
>>>
>>> I think the construct of 2 is good but have no real preference on the
>>> model/listOf issue.
>>>
>>> On 3.
>>> Given that the issue of the semantics of the identifiers has already
>>> been raised I think Solution 1 is asking for more trouble. I like
>>> solution 2 but again this needs careful thought as to how it evolves
>>> to a statement "subject has NOT predicate whose value is object" :-)
>>> This may give Neil even more sleepless nights than the current
>>> performance of the England football team.
>>>
>>> Sarah


-- 
Nicolas LE NOVERE, Computational Neurobiology, EMBL-EBI, Wellcome-Trust
Genome Campus, Hinxton CB101SD UK, Mob:+447833147074, Tel:+441223494521
Fax:468,Skype:n.lenovere,AIM:nlenovere,MSN:nle...@ho...(NOT email)
http://www.ebi.ac.uk/~lenov/, http://www.ebi.ac.uk/compneur/, @lenovere