Re: Neurolex question

[Stephen Larson]

Hi Mihail,

   Answers below:

On Sat, Dec 21, 2013 at 7:42 PM, Mihail Bota <mbo...@gmail.com> wrote:

Hi All,

I have a question about the NeuroLex structure: can a brain region A
defined in a certain species (and nomenclature) to inherit properties
from a region B, defined for a genus/family etc? It would be extremely
helpful...Also, is it possible the opposite process? A brain region
defined for all mammals to "extract", or distill, the common
properties of the regions defined in family/genus/species?
If not, I was thinking to adapt to NeuroLex what I have in BAMS for
properties inheritance of neuron types and classes.

It is an interesting proposal to try to and implement inheritence for brain regions.  NeuroLex does not natively support inheritance in that way.  However, we have been able to get knowledge from other pages to aggregate using specialized queries, so it is possible we could design something that acted like inheritance if we got more specific on what we wanted to see exactly.  Do you have an example in mind?

Best,

  Stephen

 

Mihai

--
Mihail Bota
Associate Professor of Research
Department of Neurobiology
University of Southern California

[Mihail Bota]

Hi Stephen,

I thought you already have an inheritance for connections. If you have
the relationship "part_of" for regions, then the common attributes can
be moved up.
I think the process of inheritance should be automatic. For example,
"6 layers" of the cerebral cortex has to go down to all its regions.
Here, of course, we have to be a little bit careful, because
hippocampus, considered a part of the cerebral cortex has 3 layers.
The same for regions like piriform cortex, or cortical amygdalar
regions.
Another example: layer V output neurons of the cortex are pyramidal, I think.
In the opposite direction, it would nice and useful to infer rules
like: "basal forebrain (whatever that is) sends inihibitory
connections to the cortex).
There are many examples that can be worked out. Maybe a simple one,
and complete in BAMS, would be: "retinal ganglion cells send their
axons outside of the Retina", i.e., they are projection interneurons.
Or, if anybody thinks at a more complex example, we can work that out.

What do you think of this?

And Happy New Year to everybody!

Mihai

[Stephen Larson]

On Tue, Dec 31, 2013 at 4:25 PM, Mihail Bota <mbo...@gmail.com> wrote:

I thought you already have an inheritance for connections. If you have
the relationship "part_of" for regions, then the common attributes can
be moved up.
I think the process of inheritance should be automatic. For example,
"6 layers" of the cerebral cortex has to go down to all its regions.
Here, of course, we have to be a little bit careful, because
hippocampus, considered a part of the cerebral cortex has 3 layers.
The same for regions like piriform cortex, or cortical amygdalar
regions.
Another example: layer V output neurons of the cortex are pyramidal, I think.
In the opposite direction, it would nice and useful to infer rules
like: "basal forebrain (whatever that is) sends inihibitory
connections to the cortex).
There are many examples that can be worked out. Maybe a simple one,
and complete in BAMS, would be: "retinal ganglion cells send their
axons outside of the Retina", i.e., they are projection interneurons.
Or, if anybody thinks at a more complex example, we can work that out.

What do you think of this?

And Happy New Year to everybody!

This kind of inference would be very nice to have and it is one of the features in BAMS I like the most.  It is not currently implemented in NeuroLex though, but with some clever queries could be done.  We'll put this on the list of features to implement.

Happy New Year!

Best,

  Stephen

[Chris Mungall]

Related aside: I propose that cell types such as "Neocortex X cell layer N" are renamed as cell types, e.g. "X cell of neocortex layer N". Currently the cell is named like a cell layer, but separate classes exist for the cell layers:

http://neurolex.org/wiki/Category:Neocortex_layer_5
http://neurolex.org/wiki/Category:Neocortex_layer_6

(originating in NIF_GrossAnatomy, as opposed to NIF_Cell)

For reasoning to work, the semantics of neurolex should be clear. Currently the page has:

Soma Location: Neocortex layer 5, Neocortex layer 6

I believe here the intention is that this should be interpreted as "OR" (i.e a 5/6 cell can have it's soma location in layer 5 or in layer 6, both not both. However, The actual underlying semantics in the RDF are clearly "AND". This is a problem if any kind of spatial reasoning is to be used, as nothing can be part of two spatially disjoint layers at the same time (of course, things can overlap both).

Another aside: why does clicking on "Soma Location" above take us to a different property, "located in"? In NIFSTD this is treated as a shortcut relation, the same as http://purl.obolibrary.org/obo/RO_0002100 (has soma location)

The logical properties and the textual descriptions don't seem to match - The "defining criteria" is "Basal and apical dendrites; soma in neocortical layer 5" - nothing about layer 6. Why not just rename this class as "pyramidal cell of neocortical layer 5"?

Mihail, some of the inferences you mention are possible but not straightforward. I would strongly recommend that these are encoded in OWL so that we can use standard reasoners. Note that properties are not inherited up the partonomy, unless an explicit property chain is recorded. For example, by itself "has soma property" does not propagate over part_of. However, if we encode a property chain, as we do in RO, then, this is inherited:

Property Chains:
has soma location o part of subPropertyOf has soma location

this inheritance property can itself be proven from the underlying relation definition of "has soma location" (intuitively, if a cell has its cell body in X, and X is part of Y, then the cell body is necessarily in Y)

See PMID:22402613 for details on the OWL axiomatization of these relations.

Inference of number of layers in a region is also something that only happens if you have explicit axioms that entail this (which is desirable, otherwise you end up with incorrect inferences).

Happy new year all!

--
You received this message because you are subscribed to the Google Groups "neurolex" group.
To unsubscribe from this group and stop receiving emails from it, send an email to neurolex+u...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

[Maryann Martone]

Hi Chris:

The naming scheme was proposed by Gordon Shepherd and Giorgio Ascoli, specifically so that we could generate lists of neurons that would have all cell types from the same region listed together. I know it doesn't matter for ontologies, but this was an important feature that Gordon wanted.

As per below, neocortex traditionally is considered to have 6 layers, not cerebral cortex. Although if you listen to Harvey Karten, the actual basis of the 6 layers is fanciful, leading to the need for 4a, 4b, 4c etc. But the type of inheritance you propose is exactly what we would want to implement.

Maryann

Maryann Martone, Ph. D.
Professor-in-Residence
Department of Neuroscience
University of California, San Diego
San Diego, CA 92093-0608

858-822-0745

[Giorgio Ascoli]

Chris, Maryann, and all,
I believe that the verbiage "layer 5/6 PC" in the neocortical community is indeed commonly intended to mean "either 5 or 6", but this is not necessarily a general rule. Not too long ago, in an attempt to put some order in NeuroMorpho.Org, we went through a bunch of examples of papers referring to "layer X/Y" in the literature describing the somatic location of neuronal reconstruction. We found several distinct meanings:

1. There is one neuron type with otherwise homogeneous properties, with two subclasses characterized by soma location in layers X and Y respectively;
2. In the region of interest and with the given stain, the border between two or more layers is not easy to distinguish. All the somata of the given class are assumed to be in one of particular layer, but we don't know whether that layer is X or Y;
3. The soma tend to cluster at the border between X and Y.

Unfortunately the use of hyphens and slashes become even more ambiguous with physiological terms...
All the best,
Giorgio

----------------
Giorgio Ascoli, PhD
University Professor
Founding Editor-in-Chief, Neuroinformatics
Director, Center for Neural Informatics, Structure, and Plasticity
Molecular Neuroscience Dept., Krasnow Institute for Advanced Study
MS2A1 - George Mason University, Fairfax, VA 22030-4444
Web: http://krasnow1.gmu.edu/cn3
Ph. +1(703)993-4383

[Chris Mungall]

On Mon, Jan 13, 2014 at 6:54 PM, Maryann Martone <mar...@ncmir.ucsd.edu> wrote:

Hi Chris:

The naming scheme was proposed by Gordon Shepherd and Giorgio Ascoli, specifically so that we could generate lists of neurons that would have all cell types from the same region listed together.  I know it doesn't matter for ontologies, but this was an important feature that Gordon wanted.

Ah, I see. Alphanumeric ordering of lists is still the dominant browsing paradigm.... :-)