[Data-modeling] Hormone type in the Medicine domain
Faye Li
faye at metaweb.com
Mon Apr 28 20:00:51 UTC 2008
Completely agreed. I'm all for reusing existing ontologies, especially
if a lot of experts have gone through the trouble of creating them and
have, more or less, agreed to adhere to them. I'm just realizing the
amount of great resources out there. I have borrowed a recent Unified
Medical Language System (UMLS) DVD to see what I can implement in
Freebase. Stay tuned.
Thanks,
-- Faye
Danny Hillis wrote:
> I would like to second this approach. The ontologies of biology data
> have been thought about way more than most other areas, and it is
> crucial that we take advantage of that. This approach is a little more
> complex than other freebase data models, but these kinds of details
> are important in this field.
> -Danny
>
> On Apr 26, 2008, at 1:31 PM, Dan Ruderman wrote:
>
>> I think this is a great discussion. Ben's suggestion that existing
>> ontologies be ported to the MW is an important one. By doing
>> so we will end up with a one-stop database which bridges
>> a number of currently disjoint data sets and ontologies. The
>> fact that the MW allows co-typing is central to this. The price
>> we pay is the added effort of piecing together the jigsaw puzzle.
>>
>> I might suggest that the notion of a hormone receptor might
>> best be captured as a biological function ascribed to a protein
>> (and, in turn, to its gene). In the existing data in the biology
>> domain there is an example of this. The Gene Ontology group
>> "steroid hormone receptor activity"
>> (http://www.freebase.com/view/guid/9202a8c04000641f8000000005213ec5)
>> is an annotation on a number of genes, e.g. estrogen receptor 1
>> (http://www.freebase.com/view/guid/9202a8c04000641f800000000419c5ad).
>> This annotation was made by the Gene Ontology Consortium, and
>> I doubt anyone would argue with them about it and I doubt
>> this would change. But keeping track of who said it is important
>> because
>> not everyone agrees on these annotations, and they may change
>> over time (which might be nice to track).
>>
>> So rather than simply co-typing a protein (or gene) as a hormone
>> receptor,
>> one could instead define hormone receptor as a type and link to
>> proteins
>> (or genes) though a membership type which has properties like who said
>> it, when, and what was the evidence. This follows the model that the
>> Gene Ontology Consortium itself uses. What's nice is that it allows
>> us
>> to talk about these functional annotations not as a hard fact but as a
>> statement
>> made at a given time by a given organization or person.
>>
>>
>> Regards,
>> Dan
>>
>>
>>
>>
>>
>>>> *From: *"Faye Li" <faye at metaweb.com <mailto:faye at metaweb.com>>
>>>> *Date: *April 16, 2008 12:03:52 PM PDT
>>>> *To: *"Freebase data modeling mailing list"
>>>> <data-modeling at freebase.com <mailto:data-modeling at freebase.com>>
>>>> *Subject: **Re: [Data-modeling] Hormone type in the Medicine domain*
>>>> *Reply-To: *"Freebase data modeling mailing list"
>>>> <data-modeling at freebase.com <mailto:data-modeling at freebase.com>>
>>>>
>>>> Thanks -- that was one of the things that confused me. Hormone
>>>> receptors
>>>> were defined as proteins on most online sources I consulted, but I
>>>> was
>>>> influenced in the opposite direction by the existing Gene type and
>>>> instances in the Biology domain and the muddy use of both terms in
>>>> Wikipedia. Sounds like using Protein (again, in Biology) as an
>>>> included
>>>> type would make sense here.
>>>>
>>>> The point of the explicit naming of "Human" in front of "Hormone"
>>>> is to
>>>> drive home the fact that the Medicine domain will be limited to the
>>>> scope of human instead of encompassing plant or animal. It doesn't
>>>> mean
>>>> that these hormones are limited to the human species; it does mean,
>>>> however, that hormones that don't apply to humans shouldn't be
>>>> added to
>>>> this type -- a situation that's been happening with the Disease type
>>>> that I'd like to avoid. Hormone will be linked to Disease, which has
>>>> properties like causes and treatments, both of which can be very
>>>> different depending on the species of the patient. These and other
>>>> things I want to model require that the reciprocating type be Human
>>>> Hormone.
>>>>
>>>> I could, however, create a separate type of Hormone (perhaps best
>>>> in the
>>>> Biology domain) for all hormones in the world (and add Human
>>>> Hormone as
>>>> an included type). Would that be an acceptable solution to you?
>>>>
>>>> -- Faye
>>>>
>>>>
>>>> Brian Karlak wrote:
>>>>
>>>>> On Apr 15, 2008, at 5:07 PM, Faye Li wrote:
>>>>>
>>>>>> Also, is a Hormone Receptor always or pretty much always a Gene?
>>>>>>
>>>> We have
>>>>
>>>>>> a Gene type in the Biology domain that I'd be more than happy to
>>>>>>
>>>> add as
>>>>
>>>>>> an Included Type if it makes sense. Medical review of the type and
>>>>>> property names would also be appreciated.
>>>>>>
>>>>> Hi Faye --
>>>>>
>>>>> Hormone Receptors are all proteins, which in turn are encoded for
>>>>> by
>>>>> genes. However, the distinction is often blurred in common usage
>>>>> since there is a rough one-to-one correspondence between a gene and
>>>>> and its protein product. Furthermore, genes often share the same
>>>>> name
>>>>> as their protein product.
>>>>>
>>>>> Even though common usage often confuses the two concepts, it would
>>>>> probably be incorrect to cotype all Hormone Receptors as Genes. It
>>>>> would be more correct to cotype them as Proteins.
>>>>>
>>>>> Unfortunately, it seems that Wikipedia usually confuses the two
>>>>> concepts in their articles. For instance, the A2a Receptor
>>>>>
>>>>>
>>>> <http://sandbox.freebase.com/view/guid/9202a8c04000641f8000000006bd15e1
>>>>
>>>> blurb
>>>>
>>>>> starts with:
>>>>>
>>>>> The adenosine A receptor, also known as ADORA2A, is an adenosine
>>>>> receptor, but also denotes the human gene encoding it.
>>>>>
>>>>>
>>>>> It appears that we have continued this conflation of Gene and
>>>>> Protein
>>>>> in freebase. I see instances where we (automatically) type these
>>>>> entries as a Gene. This irks the bio geek in me. For common use
>>>>> this
>>>>> will work, but any biological application that tries to use this
>>>>> info
>>>>> will get tangled up pretty quick.
>>>>>
>>>>> Because of this, I'd recommend against the automatic cotyping of
>>>>> Hormone Receptor with Gene, unless we're comfortable propagating
>>>>> this
>>>>> issue.
>>>>>
>>>>> One final suggestion: you should probably name your type "Hormone",
>>>>> instead of "Human Hormone". Most (if not all) of your entries are
>>>>> hormones in a wide variety of species. Dogs, mice and deer all
>>>>> produce prolactin and insulin, all for pretty much the same
>>>>> purpose.
>>>>>
>>>>> Brian
>>>>>
>>>>>
>>>>>
>>>> ------------------------------------------------------------------------
>>>>
>>>>> _______________________________________________
>>>>> Data-modeling mailing list
>>>>> Data-modeling at freebase.com <mailto:Data-modeling at freebase.com>
>>>>> http://lists.freebase.com/mailman/listinfo/data-modeling
>>>>>
>>>>>
>>>> _______________________________________________
>>>> Data-modeling mailing list
>>>> Data-modeling at freebase.com <mailto:Data-modeling at freebase.com>
>>>> http://lists.freebase.com/mailman/listinfo/data-modeling
>>>>
>>>>
>> _______________________________________________
>> Data-modeling mailing list
>> Data-modeling at freebase.com
>> http://lists.freebase.com/mailman/listinfo/data-modeling
>>
>
> _______________________________________________
> Data-modeling mailing list
> Data-modeling at freebase.com
> http://lists.freebase.com/mailman/listinfo/data-modeling
>
>
More information about the Data-modeling
mailing list