Re: cell types, brain regions mentioned in gensat

Thanks, Mihail.

I was hoping more experts would step in and provide some useful  
guidance here.

Are all the ontologies you refer to below - neurons in particular -  
publicly available in OWL, RDF, or, at least, according to some XML  
Schema?

The crux of the problem we run into here is the fact that the  
terminologies must be usable to semantically integrate across data  
sets - just mouse data sets - but those data sets don't all use the  
same nomenclature for neurons OR for brain regions.  For region  
terms, as both you and I pointed out, sometimes they don't refer to  
the same geometric regions and/or those regions are not specified in  
registered coordinate spaces.

Cheers,
Bill

On Mar 2, 2007, at 2:34 PM, Mihail Bota wrote:

> It is very difficult, actually not very advisable, to borrow  
> nomenclature
> characteristic of one species to another species, without
> similarities/homologies discussion. For example, in rodents there  
> is no
> separation between caudate and putamen (therefore the term  
> caudoputamen),
> while in humans and monkeys these structures are separate. Thus, for
> rodents, BAMS has several nomenclatures, and it can be easily used.  
> ABA
> was actually made after Swanson 1998/2004, rat.
> Caudoputamen is a part of striatum (sometimes called the dorsal part,
> to make the difference of the ventral part, which is the nucleus  
> accumbens),
> therefore is not a "CNS unspecified".
> The direct link to caudoputamen in BAMS, where you'll find
> its tree (also specified in the Swanson-1998 XML page) is this:
> http://brancusi.usc.edu/bkms/brain/show-braing2.php?aidi=129
> Please look at the definitions associated to each of the regions  
> names in BAMS
> for details in defining those.
>
> Regarding cell ontologies, you may want to check BAMS Neurons  
> Ontology,
> which contains about 250 terms used for neurons in different  
> regions of
> the rat brain. There are about 37 brain regions "filled" with  
> neurons in
> BAMS, with retina and the cerebellar cortex being complete, and other
> regions with high degree of coverage (paraventricular nucleus of  
> hypothalamus for
> example). The approach is rather different of Cell Ontology and  
> Senselab,
> because we try to specify the classification criteria, as well as
> other metadata such as references, collators, annotations, etc.
> We register neuron types and classes defined by a large set of  
> techniques
> and strive to insert relations between terms.
> You can use the ontology from the "Cells" link of the BAMS menu. The
> neurons concepts are also in XML pages, more specifically in the  
> Swanson
> 1998 page.
>
> CNU stands for "Cerebral nuclei" and is defined in rat nomenclature
> Swanson 2004 (also used by ABA), and not for "unspecified". Please  
> check
> BAMS for its definition
> (http://brancusi.usc.edu/bkms/brain/show-braing2.php?aidi=12109 ,  
> click on
> button "More" for its definition), or please find it in XML page
> Swanson-2004.
> Paxinos/Franklin nomenclature does not have this term. Instead it  
> has "forebrain",
> "midbrain", "hindbrain". The hierarchical organization of Paxinos/ 
> Franklin
> was made by us, and therefore you will not find it in the atlas.
>
> Mihai
>
> On Fri, 2 Mar 2007, William Bug wrote:
>
>>
>> 	+++++++=========== 	NEURONAL CELL TYPES 	===========+++++++
>>
>> Chris is correct - the MOST LIKELY shared community ontology to
>> address the issue of cells would be the Cell Ontology, though in its
>> currently distributed form - see (http://www.berkeleybop.org/
>> ontologies/#mappings) - it's very problematically both ontologically
>> and in terms of what a neuroanatomist and cellular neurobiologist
>> would want.
>>
>> This brings us to a particular problem with the GENSAT AND the
>> BrainPharm/SenseLab cell types - relating to shared cell type
>> ontologies.  It's POSSIBLE the GENSAT folks started with the Cell
>> Ontology.  Some of their cell terms are in there.  It's even more
>> likely this is something they grew themselves.  This is most
>> definitely true for many of the cell "sub types".
>>
>> The Cell-Centered Database (CCDB) has done A LOT of work on this
>> issue.  It would really be a terrible waste if this is not being
>> included in whatever work is going on amongst the CL Ontology folks
>> and GMOD folks - at least in the domain of neuronal cells.  It's
>> quite likely some form of the CCDB work is available in OWL.  I've
>> cc'd Maryann & Stephen who've been working on CCDB ontologies.  They
>> could answer that question.
>>
>> With BrainPharm - at least as I understand it (Luis Marenco and/or
>> Kei of Yale would be best suited to answer this) the "cell types"
>> they create have been based both on the wealth of knowledge Gordon
>> Shepherd brings to this issue of neuronal cell types (at least as a
>> core organizing principle).  This - possibly with input from some
>> shared, distilled knowledge resources such as MeSH (as Don says) has
>> then been built out by the 1000s of cellular neurobiologists who've
>> contributed to the the SenseLab database - very empirically.  To my
>> knowledge (Kei & Luis?), they have not tied this into any current,
>> shared community ontology for cell types - and I'm certain there are
>> real reasons why this is the case.
>>
>>
>> 	+++++++=========== 	RODENT BRAIN ATLASES and BRAIN REGION
>> SPECIFICATION 	===========+++++++
>>
>> As for brain regions, this TOO is problematic.  What Gwen states is
>> colloquially correct - you will find brain regions specified in
>> atlases.  However, primate atlases and rodent atlases do not
>> correspond in many areas (this is an ACTIVE area of research and a
>> major task for us in BIRN given our mandate to integrate across
>> animal models and human clinical neuroimages related to the diagnosis
>> and study of neurodegenerative disease).
>>
>> Limiting just one sub-species - say the Mus musculus musculus x Mus
>> musculus domesticus hybrids that gave rise to most of the inbred
>> strains (C57Bl/6, DBA/2, etc.) commonly used to create mouse models
>> of disease - recombinant inbred, transgenics, congenics, etc. - there
>> are AT LEAST 3 MRI/MRM atlases of the C57Bl/6 brain:
>> 	LONI - http://www.loni.ucla.edu/MAP/Atlas/
>> 	Ma et al. - http://www.bnl.gov/ctn/mouse/
>> 	Mouse Imaging Center (Toronto) - http://www.mouseimaging.ca/ 
>> research/
>> var_brain_atlas.html
>>
>> and AT LEAST 2 histological atlases of the C57Bl/6 brain:
>> 	LONI - http://www.loni.ucla.edu/MAP/Atlas/Atlas_Archive.html
>> 	NeuroTerrain - http://fireball.drexelmed.edu/neuroterrain/ 
>> tools.shtml
>>
>> MRI/MRM atlases are just starting to get under 100 microns in
>> resolution (compared to histologic resolution generally 20 - 50
>> microns 3D isotropic) and don't have nearly comparable contrast (yet)
>> to histological atlases required to perform highly granular region
>> segmentation.  This contrast issue is especially true in the deep
>> regions of the diencephalon where many regions important to
>> neurodegenerative disease reside (s. nigra, n. basalis, striatum,  
>> etc.).
>>
>> And - guess what - they don't all share the same "terminology" for
>> naming those regions they do delineate (see BAMS links below) NOR are
>> the regions (the 3D objects) they delineate for a given region -
>> e.g., "hippocampal formation" -  geometrically equivalent - NOR do
>> they place these geometries in the same 3D coordinate system.
>>
>> I realize it must seem I'm getting into too much technical detail
>> here - BUT- IF the goal is to make MEANINGFUL connections via SPARQL
>> queries, this sort of behind the scenes "sausage making" issues are
>> very important issues.
>>
>> For primate brain regions, the SMARTEST source for a shared
>> "terminology" is NeuroNames.  As I've mentioned before, this MIGHT be
>> available in RDF or OWL based on work the BrainInfo/NN PI - Doug
>> Bowden has done with Jack Park at SRI (I've cc'd Jack on this
>> email).  Many names from other species are in there (though not in
>> the downloadable versions of NN which are CIRCA 2002).  There is a
>> knowledgeable - the Brain Architecture Managment System (BAMS) from
>> Larry Swanson's lab which is curated by Mihail Bota - which makes
>> available in XML format (don't know what the XSD is) the brain region
>> specifications for several STANDARD rodent brain atlases.  You'll
>> find these by navigating from the BAMS home page (http://
>> brancusi.usc.edu/bkms/) to the 'XML' link:
>> 	http://brancusi.usc.edu/bkms/bamsxml.html
>>
>> I would note the ABA uses the Franklin & Paxinos altas (2001).
>>
>> HOWEVER - there are many regions in your spread sheet that are not in
>> the BAMS Franklin&Paxinos download - for instance "CNU" - which may
>> actually be a home-grown term for "CNS - unspecified".  Isn't that
>> special.  Here's why that might be.  Go to the NN/Braininfo page for
>> one of the regions the ABA classifies as "CNU" such as striatum:
>> 	http://braininfo.rprc.washington.edu/Scripts/
>> hiercentraldirectory.aspx?ID=207
>> Click on the "Where is it in the brain?" link.
>> You will note Striatum, Globus Pallidus, and Amygdala are listed as
>> being in the "Basal Ganglia".  Leaving the Amygdala aside, certainly
>> in humans, the Striatum & Globus Pallidus (two regions critical to
>> the PD Use case, BTW) are considered to be in the Basal Ganglia.
>> In rodents - especially in mice - this classification is very
>> problematic.  You won't get any two mouse neuroanatomists to TRULY
>> agree on all the details, however, there is a generally accepted
>> region that sort of subsumes the same entity (based on various
>> histological means of defining structural boundaries as well as other
>> means such as structure-function, phylogenetic, and ontogenic
>> relations) called the "caudoputamen" by most (if you drill into
>> Striatum on BrainInfo, you'll find the Striatum has_part Caudate
>> Nucleus.  Caudoputamen IS NOT equivalent to the "Basal Ganglia", but
>> it is probably correct to say it is ANALOGOUS to what in a primate
>> would be the Putamen and the Caudate Nucleus.
>> ABA may just have thrown up there hands and said - given their high-
>> throughput directive - no one seems to agree on this issue, so lets
>> be careful not to make an assertion that entails committing to an
>> issue where there is disagreement.  Let's just call Striatum "CNS
>> Unspecified".
>> Having said this, there is a Caudoputamen in ABA and it's classified
>> as "Striatum dorsal Region (STRd).
>>
>> I would guess Mihail, Maryann, and others would be able to give you a
>> better explanation of the how to address these difficult issues of
>> anatomical region correspondence between rodent brain image sets
>> (GENSAT, ABA) and atlases (Franklin & Paxinos, NeuroTerrain, Ma et
>> al., etc.).
>>
>> Hope this helps some.
>>
>> Cheers,
>> Bill
>>
>>
>> On Mar 2, 2007, at 11:47 AM, Chris Mungall wrote:
>>
>>>
>>>
>>> Umm, the OBO Cell ontology? There are a few ontological issues with
>>> OBO-Cell right now, but these are currently being addressed. There
>>> are also efforts within OBO to link cells with the brain regions
>>> they are part of, although these are currently focused on model
>>> organisms.
>>>
>>> On Mar 2, 2007, at 8:13 AM, Donald Doherty wrote:
>>>
>>>>
>>>> Alan,
>>>>
>>>> The region names are all available in the MeSH...would that give
>>>> you the
>>>> taxonomy you need? I don't know of a similar source for cell types.
>>>>
>>>> Don
>>>>
>>>> -----Original Message-----
>>>> From: Alan Ruttenberg [mailto:alanruttenberg@gmail.com]
>>>> Sent: Friday, March 02, 2007 3:21 AM
>>>> To: Bill Bug; kc28 Cheung; June Kinoshita; Gwen Wong; Donald  
>>>> Doherty
>>>> Cc: public-semweb-lifesci@w3.org
>>>> Subject: cell types, brain regions mentioned in gensat
>>>>
>>>> I'm making progress in converting gensat to rdf.
>>>>
>>>> For mapping considerations, here is the list of cell types  
>>>> mentioned
>>>> in gensat, followed by the list of brain regions. If we are  
>>>> going to
>>>> do cross queries we will need to find standard names for these.  
>>>> Bill,
>>>> are these classes in birnlex? If not, we need to spawn a task to
>>>> identify a vocabulary we will use for these.
>>>>
>>>> Note that we get a region<->neuron association via gensat where  
>>>> they
>>>> annotation both a region and a cell type.
>>>> Note also some amusements, like the presence of lung as region  
>>>> in an
>>>> ostensibly CNS database.
>>>>
>>>> I've also attached the "ontology.csv" from the Allen Brain Explorer
>>>> application, which I presume gives their hierarchy of brain  
>>>> regions/
>>>> subregions. I've put labels on the first 3 columns which I think
>>>> encode the hierarchy.
>>>>
>>>> The other interesting annotations, are the gene, the location,
>>>> orientation, and size of the image, as well as some broad  
>>>> categories
>>>> of qualitative expression, such as whether it is localized of  
>>>> widely
>>>> expressed. There is also gender and a few categories of age.
>>>>
>>>> There are ~60K images in gensat.
>>>>
>>>> BTW, if someone has a theory of what the other number in  
>>>> ontology.xls
>>>> are, I'm all ears.
>>>>
>>>> -Alan
>>>>
>>>>
>>>>
>>>>
>>>>
>>>
>>>
>>
>> Bill Bug
>> Senior Research Analyst/Ontological Engineer
>>
>> Laboratory for Bioimaging  & Anatomical Informatics
>> www.neuroterrain.org
>> Department of Neurobiology & Anatomy
>> Drexel University College of Medicine
>> 2900 Queen Lane
>> Philadelphia, PA    19129
>> 215 991 8430 (ph)
>> 610 457 0443 (mobile)
>> 215 843 9367 (fax)
>>
>>
>> Please Note: I now have a new email - William.Bug@DrexelMed.edu
>>
>>
>>
>>
>>
>
>
>

Bill Bug
Senior Research Analyst/Ontological Engineer

Laboratory for Bioimaging  & Anatomical Informatics
www.neuroterrain.org
Department of Neurobiology & Anatomy
Drexel University College of Medicine
2900 Queen Lane
Philadelphia, PA    19129
215 991 8430 (ph)
610 457 0443 (mobile)
215 843 9367 (fax)


Please Note: I now have a new email - William.Bug@DrexelMed.edu