August 18, 2009
1) Current state of Neurolex vs NIF ontology
- Fahim has created the Neuon to molecule and Neuron to brain region bridge files, taking the information provided in Neurolex and creating the necessary restrictions in OWL.
- Bridge file location: http://ontology.neuinfo.org/NIF/BiomaterialEntities/NIF-Neuron-NT-Bridge.owl
- Generated inferred classification of neurons by neurotransmitter is saved at: http://ontology.neuinfo.org/NIF/BiomaterialEntities/NIF-Neuron-NT-Bridge_Inferred.owl
- Reclassification of neurons according to neurotransmitter and brain region will be available for search through NIF through the creation of classes for, e.g., GABAergic neuron. Selection of this term from the NIF autocomplete function in the NIF search box will expand to a list of neurons that are classified automatically according to neurotransmitter. Similar classes will be created for brain regions, e.g., Hippocampal neuron.
- We are considering down the line how we would want to make this search function available through NIF beyond the simple search box.
- Maryann has been removing duplicate classes from the Neurolex Wiki site; still work to be done in spinal cord.
2) Macrorelationships: For the time being, NIF will use so called macrorelationships to relate classes where to use an OBO RO property would require two or more relationships. This convention is adopted both for simplicity in user interfaces and for computational simplicity. It is NIF's stated intention to define a property chain using RO relationships for each of these custom relations. The current custom macrorelation in use is "has neurotransmitter". The "has soma location" will be dropped in favor of "has location", an RO property
3) Naming conventions: We discussed Giorgio and Sridevi's proposal and other issues that have arisen in applying our current naming convention. The following modifications/additions were made:
- Use of capital letters vs lower case in the neuron category names. We propose to capitalize the first letter of the brain region while keeping the following terms lower case (e.g. Dentate gyrus granule cell). Exceptions: (i) terms named after an investigator should be capitalized (e.g. Cerebellum Purkinje cell); (ii) acronym terms are all capitalized (e.g., Dentate gyrus HICAP cell). We recognize that some in the ontology community do not favor this style, but as it is enforced by our Neurolex Wiki, we will use it for now.
- Primary and secondary brain regions: Each neuron will be designated by its primary brain region first, e.g. Cochlear nucleus followed by a secondary brain region qualifer, e.g., dorsal, followed by a functional, biochemical or morphological property that distinguishes the cell class or a proper name, e.g., Cochlear nucleus ventral octopus neuron.
- For neurons that are found in two or more brain regions that cannot be related clearly through a part of relationship, e.g., the large multipolar neuron of the endopiriform nucleus and piriform cortex, we will create two classes: Endopiriform large multipolar neuron and "Piriform cortex deep multipolar neuron". These may both be children of a superclass, "Large multipolar neuron" that will be generated through an inferred hierarchy based on properties, but will not be asserted.
- Cross products: NIF will create cross products between two brain region parcellation schemes to identify a specific location, e.g., stratum oriens of hippocampal area CA3. The preferred label for these cross products will be the major brain region followed by the subregion, e.g., CA3 stratum oriens, meaning the stratum oriens part of area CA3. Other forms will be assigned as synonyms.
- Use of special symbols: Because special symbols, e.g., parentheses, commas and slashes, cause problems with some information systems and can introduce ambiguity into use of classes in papers, we will avoid their use. This convention will pose a problem sometimes for assigning clear unambiguous labels to brain regions preceding cell types.
- Cell vs neuron: All classes of neurons will end in either neuron or cell, rather than any other qualifier. When possible, we will use neuron because it provides more information. In cases where the common name of the class uses "cell" instead of neuron, we will retain the use of cell. If, however, both cell and neuron are used, even if cell is used more frequently, we will use neuron as the label. This convention will require that we change the preferred label of many neuron classes. The "cell" variant should be listed as a synonym.
- Sometimes we use the term "interneuron" as a distinguishing characteristic, e.g., substantia nigra GABA interneuron. For now, we will accept "interneuron" as an acceptable variant on neuron, but again, if there is a choice, we should standardize to neuron.
- Acronyms: When neuron category names consist of acronyms, these should be spelled out in the definition field with the appropriate letters capitalized (as in "HIlarCommissural-Associational Pathway related cells" for HICAP).
- Properties: See above policy on the use of macrorelations. Current properties proposed for interface and macrorelations
- We will replace the "has part" with "located in" for relating cell class to brain region
- We will use the "has neurotransmitter" relation for relating neuron to neurotransmitter
- We will use "has neurochemical marker" for relating neuron to distinguishing chemical markers of neurons.
- We will use "has molecule" as a general way to add molecules to a cell class that are not necessarily a distinguishing characteristic
- Assigning location: We will use the location of the neuron soma and the bulk of the dendritic tree as the criteria for assigning location to a brain region. The previous policy was to use the cell soma and assign brain region granularity according to where the soma was likely to be. Giorgio brought up that for many interneuron classes, the soma location is less important and characteristic than the dendritic tree.
- We will still assign locations to the dendrites and soma, but as separate parts in the form: Neuron X has part soma; Soma of Neuron X has location Brain region Y. Ideally, the interface will help with this so that when one assigns a soma as part of a neuron, the cross product of Neuron class X subpart is automatically generated so that users do not have to use multiple steps. This facility is not yet implemented in any of our user tools.