The NIF cell working group is composed of Gordon Shepherd, Perry Miller, Kei Cheung, Maryann Martone, Giorgio Ascoli, Sridevi Polavaram and Fahim Imam. They will be working closely with the INCF to develop standards for naming and describing neurons. As a specific NIF goal, they will be fleshing out the NIF cell list and creating a bridge file that relates cell type to brain region. The group meets regularly by phone.

At the first meeting, it was decided that Giorgio and Sridevi would concentrate on hippocampal neuron subtypes and describing their differentiating properties. Giorgio put together a document that outlines the general issues. The goal is to flesh out the template that is posted on the Neurolex. Sridevi is working with Stephen Larson to get up to speed.

3/12/09 meeting: Gordon made a proposal for standardizing the names of all neurons in the current Neurolex, using the brain region as the first term. The group has agreed that including the brain region as the main differentiating criteria is a reasonable way to start and allows Fahim to construct the underlying logical definitions in OWL in a very straightforward manner. He has gone through the entire list of cells and made recommendations for their names. The proposal is available here. The rest of the working group will be going through the list and making their recommendations. Once the list is finalized, Fahim will be swapping out the current NIF label with the standardized label. Sridevi also recommended that we add an annotation property: Common name to distiguish the usual convention from other synonyms and variants. The group concurred.

The group also agreed that the brain region that will be assigned to the cell class represents the location of the cell soma, but that other parts of cells an be located in different brain regions or subregions. The cell soma location will be assigned at whatever level of granularity is required to distinguish two subclasses of cell, e.g., if there is no difference between Chandalier cells in CA3 stratum radiatum vs CA3 stratum lucidum, then the brain location is assigned as CA3. We also know that brain region is only one property by which we distinguish neurons from one another and that the list of properties assigned will be critical to creating the necessary ability to reclassify according to multiple criteria.  NIFSTD does try to adhere to the OBO best practice of minimizing the number of custom properties and using the relationss ontology (RO) where possible.  However, that currently presents an additional computational overhead for computing relationships in cases where multiple properties need to be used to express the relationship between classes.  RO is still also under development and adding additional properties takes time.  In the meantime, we will take advantage of the OWL 2.0 capability to use macrorelations to express property chains.  Ideally we will agree to use the same set of macro relations as the other groups working on neuron ontologies and will eventually express these as sets of RO property chains.  The current proposed set of properties, some of which represent macrorelations that can be expressed as multiple OBO relations (in red), are as follows:

Cell to brain region:  "has soma location"/"soma located in" "cell has part soma;  soma is part of brain region"

Cell to molecule:  "has part" ;  Neuron "has part" molecule

Neuron to neurotransmitter:  "has neurotransmitter"/"neurotransmitter of":  Neuron "has part" molecule and molecule "has role" neurotransmitter.  We're working with others to express this property in terms of other relations from relations ontology (ro) while maintaining the simplified macro one through property chain.

The relationship among Neurolex Wiki, NIFSTD and Wiki Neuron was discussed. We have agreed that the Neurolex wiki is the entry point where communities, including the cell working group, can contribute content and work through issues. The content contributed to Neurolex gets formalized into the NIFSTD OWL files which get served through Ontoquest. Fahim recommended that we create a NIF Cell account for the wiki so that it easier to track the changes made to the Wiki by the cell group. Because we want to be able to track contributions by individual users, we decided against the group account but we will look into Wiki tools for creating groups.

However, we identified the need for a place where the current, blessed version of the NIFSTD cell list and associated properties are easily accessible, i.e., don't require Protege to read. The problem with the current Neurolex wiki is that it doesn't translate OWL properties particularly well, so if we go from the NIFSTD OWL file back to Neurolex Wiki, we lose work. Perry and Gordon may have an existing solution that will work for this so that we can provide interested neuroscientists with a list of NIF neurons in XML or RDF format or in an Excel spreadsheet or, at the least, in a browseable web hierarchy.

March 2009: Gordon has compiled a list of cell types from Cajal (as translated by Swanson). I have attached the list to this page.

July 2009: Notes from the Cell Working group meeting at the NIF AHM

August 2009:  Minutes from Cell working group meeting

January 21, 2010:  Discussions about granularity at the species and brain region level.  Most neuron properties, e.g., cell size, cannot be described generically but must be assigned at the species level.  Gordon is working on filling out individual properties for different species.  We propose putting these classes, e.g., rat cerebellum Purkinje cell, as children of "Cerebellum Purkinje cell" with a logical restriction, i.e., a rat cerebllum Purkinje cell is a cerebellum Purkinje cell from the rat, so that it is built up from Neurolex core classes.  A similar problem arises with subregions.  Many researchers feel that a pyramidal cell in somatosensory cortex is not the same class as a pyramidal cell in prefrontal cortex.  But this leads to a combinatorial problem, particularly since there is no standard way of describing cortical regions. 

The basic information structure has no problem with these different ways of classifications as they are all computable from the properties.  However, the Neurolex Wiki has to think about presentation so that someone coming to the Wiki can understand what classes are there and how to generate other classes from them.  Therefore a completely flat list combining all these axes would lead to too much confusion if we were to expose all of these classes through the Wiki.  Giorgio feels, though, that there are few enough deep characterizations of cortical neurons from restricted regions of the cortex that the problem will not be so large that we can't handle it.  So he is going to take a stab at it.

Paper:  Giorgio presented an outline of a paper that would come from this work and possibly the work of the Neuronal Registry and INCF task forces.

August 3, 2010: 

Present:  Gordon, Giorgio, Sridevi, Kei, Maryann, Stephen, Fahim

    1. Comments from Xwiki

KarenSkinner says:
(1)Would someone please post a several use cases illustrating how the "naive" user would use this sort of classification in making searches in NIF?
(2) Is this the convention now used by most neuroscientists, such as electrophysiologists or those doing neurogenetics? How easy/hard will it be to map legacy data sources to this ontology when federating them in NIF?
(3) How does this convention relate to vocabularies for describing diseases or studying diseases? For example, if an investigator wanted to search across databases for data related to dopamine releasing cells, would they now need to make multiple queries for the dopamine cells in each brain region?

-2009-03-26 11:57:41.537

2/16/2010: Attached email from Doug Bowden to Gordon Shepherd. Document includes Doug Bowden's responses to a list of questions Gordon Shepherd proposed on February 10th, 2010.

5/21/2010:  Maryann, Stephen, Sarah, Giorgio, Gordon, Kei, Sridevi:

Stephen said that they've gotten an export of the Neurolex wiki up as a SPARQL endpoint.  NIF also has a SPARQL endpoint for the NIFSTD.  There are some interesting differences between the RDF graphs generated from these two sources.  Stephen will send out the links for the Linked Data.