Difference between revisions of "Hirudo ontology"
|Line 158:||Line 158:|
Latest revision as of 04:12, 19 January 2009
To create a knowledgebase of cells and synapses in Hirudo in Neuronbank, analogous to the Tritonia branch, the first step is to create an ontology of Hirudo cells. NeuronBank has a core ontology that is common to all Branches. In addition, each branch can be customized with an extensible ontology.
The extensible ontology for Tritonia cannot be directly mapped to Hirudo, primarily because the leech has repeated ganglia, and generating one copy of each cell for each ganglion seems inefficient. The purpose of this article is to propose an ontology for Hirudo cells that will be convenient for those entering the data, as well as for those searching for information in the knowledgebase. The author, Daniel Wagenaar, hopes that other leech researchers will contribute to this page, refining the proposal until it will work for everybody.
This document consists of the following sections: Tritonia Ontology: a summary of the ontology and nomenclature used in the Tritonia branch of NeuronBank; Usage Scenarios: examples of how I think this knowledgebase could be used; Examplar Cells: examples of cells that have progressively more complex ganglionic specializations; and finally, Proposed Hirudo Ontology: the actual proposal. It should be noted that this document is a work in progress, and contributions are invited, especially if you think of additional usage scenarios, or if my Examplar Cells do not cover the intricacies of your favorite neuron.
The cells in the Tritonia branch of NeuronBank are identified by their Name, and an automatically generated AccessionID. Other properties that may be defined for a cell are: Cell count (how many of these cells are there?) Soma coloration, Swim phase (when is this cell active during swimming?), Molecule (neurotransmitter produced), Nerve projection (to which ganglion does this cell project), Resting activity (spontaneous firing, e.g. bursting, irregular, etc.), Circuit (which circuit is this cell part of, e.g. Swim CPG), and Soma diameter. In addition, under Outputs, one can specify to which other cells this neuron sends axons, and whether the synapse is chemical or electrical. There is also a corresponding Inputs section; I do not know whether these are automatically generated. (Paul?) Finally, arbitrary comments and literature references can be placed in the My annotations section.
The following are sketches of questions one might ask of the Hirudo NeuronBank. If this system is going to be useful, the ontology should be such that these questions can be asked conveniently, and that the user can be confident the answer is complete in that all pertinent information present in the system is returned.
Alice wants to know which cells in a midbody ganglion produce serotonin.
Bob wants to know whether the Retzius cell in ganglion 7 is in any way special.
Claire is worried that the HN cell she impaled in ganglion 3 has a poor resting potential, and wants to know its typical values.
Daniel has found a cell that excites the DE-3 cell in the next posterior ganglion, and wants to know which known cells have such a projection.
Please add your own scenarios here!
Tr1 -- This is a cell in the head brain of the leech that triggers swimming. It is a simple case, because it occurs in only one ganglion, and not in any midbody ganglia. For Tr1, specifying its location as "Head brain", or "Subesophageal ganglion" would be sufficient.
DE-3 -- This cell, which excites the dorsal longitudinal muscles, is one step more complex, because it occurs in the midbody ganglia. Fortunately, cells DE-3 in each midbody ganglion appear to be almost exact copies of each other, so specifying its location as "Midbody ganglia", is sufficient, if we also note that it occurs "Bilaterally" and in ganglia "1 through 21".
'HN' -- These cells, which control the heart rhythm, are more demanding yet. It is not enough to note that they occur "Bilaterally" and in ganglia "1 through 7": it is important to realize that HN cells in different ganglia have different projections: HN(1, 2) do not project to any HE cells, but they do inhibit HN(3, 4) ipsilaterally. HN(3) projects to all HE cells, inhibits HN(1, 2, and 5) ipsilaterally, and inhibits HN(3, 4) contralaterally. HN(4) is much like HN(3), but projects only to HE cells in ganglion 5 and up. HN(5) does not project to HE cells, but inhibits HN(6, 7) bilaterally. Finally, HN(6, 7) project to all posterior HE neurons. (Kristan, Calabrese, and Friesen, 2005: "Neuronal control of leech behavior", Prog. Neurobiol. 76(5) 279-327.)
I have listed these connections exhaustively to make several points:
(1) All HN cells have a lot in common;
(2) Ganglionic specializations may apply to one specific ganglion, or to several ganglia, and tend to affect only some of a neuron's properties.
(3) In some cases, it is desirable to identify projections by an absolute ganglion number, sometimes a relative indicator is more useful.
Rz -- The well-known Retzius cell occurs both in the brain ganglia and in midbody ganglia. Moreover, Rz(5) and Rz(6) are different from the other midbody Rz cells.
Proposed Hirudo Ontology
All the above properties in the Tritonia branch seem useful, except perhaps for Soma coloration. If we are going to have Swim phase, we probably should also have Crawl phase. There should be a distinction between a cell that is known to not oscillate with swimming (Swim phase: None) and a cell for which the involvement in swimming is not known (Swim phase left unspecified).
Specification of cell location is the main goal of this document. At the highest level, we want to say whether it occurs in the Head brain, in the Midbody ganglia, or in the Tail brain. (Can someone who works on the brains refine those into ganglia if that seems useful?) As examplified by the Rz cell, it is possible for a neuron to occur both in brain ganglia and in midbody ganglia. For cells that occur in midbody ganglia, we should be able to specify a range of ganglia in which they occur, e.g. "1-21", "10-18", or "1-4, 7-21".
Cells that are specialized in certain ganglia, should have one generic entry, named, e.g., "Rz", or "HN". In this generic entry, all properties that apply to all instances of the neuron are defined, as well as properties that apply to "most" instances. The generic entry will contain references to Specializations. For Rz, specializations will define "Rz(head brain)", and "Rz(5,6)". For HN, specializations will define "HN(1,2)", "HN(3,4)", "HN(5)", and "HN(6,7)". In the entries for the specializations, only those properties that differ from the generic cell are specified. Thus, in the entry for Rz(5,6), we do not say again that it is a serotonergic cell. The reason for not repeating information, is that if some new piece of knowledge about the Rz cells is added later on, we do not want to force the person adding that information to also update all the specializations.
For projections from midbody neurons to other midbody neurons, we need to specify (1) whether the projection is ipsi-, contra-, or bilateral, and (2) whether it projects within the same ganglion or to a range of ganglia. Ranges are to be specified relative to the source, with -1 referring to the next more anterior ganglion, and +2 to the second more posterior ganglion. For specializations (e.g. "HN(5)"), it it often more natural to specify absolute ganglion numbers, so perhaps we should have one property named TargetStyle which can have values intraganglionic, relative range, and absolute range. For the latter two values, additional properties AnteriormostTarget and PosteriormostTarget specify the beginning and end of the range. I think that it is sufficient if clicking on the target neuron name takes the user to the generic page for that neuron, because it is easy enough to navigate to the relevant specialization from there.
Note that there is nothing that prevents specifying connections between specialized neurons. For instance, it is just as easy to specify an inhibitory synapse from "HN(5)" to "HN(6,7), bilateral" as it is to specify an inhibitory synapse from "DI-4" to "DI-3, ipsilateral" or from "208" to "DE-3(+1), bilateral".
Note also that if it is later discovered that HN(6) and HN(7) are actually somewhat different, it should be straightforward to define "HN(6)" and "HN(7)" as specializations of "HN(6,7)". Creating such a hierarchy of specializations has the advantage that common features of HN(6) and HN(7) can be stored in the "HN(6,7)" node, again aiding in maintainability.
Suggestions by Brian Burrell
(Following are lightly edited excerpts from an email by BB to DW, December 2007.)
Some interneurons are not part of a specific circuit (or are not obligatory parts of a circuit), but instead are modulatory cells that likely affect many behaviors/circuits. The Rz and Lidia Szczupak's non-spiking inhibitory neuron come to mind. There should probably be an option to categorize neurons based on their neuromodulatory role.
There is lots of modeling in the leech community and we may want to be able to include that sort of information and links to where different models can be accessed, e.g. Model DB.
Suggestions by Ron Calabrese
(Following are edited excerpts from an email by RC to DW, December 2007.)
For the HN neurons things are difficult as DW points out, and ultimately only by using the ontology can we figure whether it will need some modification. But even at this point the ontology might be modified. The general properties of HN neurons are very few in my opinion so I (RC) will pass them by you for your thoughts about whether I'm missing your point. Bilateral interneurons occurring midbody ganglia 1,2,3,4,5,6,7,15,16 (note the large gap 7 to 15) no other properties are held in common. E.g., while we are confident that the HN(3) and HN(4) neurons are cholinergic and it is likely that the HN(6) and HN(7) neurons are also we must remain agnostic about the HN(1), HN(2), HN(5), HN(15) and HN(16) neurons (especially the first three whose properties are so different from the HN(3) neurons, e.g. Also neurons like the HN(3) and HN(4) neurons we often lump together but this is not really true because the HN(3) neuron inhibits it local motor neuron and excites the ipsilateral HN(4) while the HN(4) neuron does not.
Even the HE neurons which appear so homogeneous are all different in terms of the presynaptic inputs and probably in terms of intrinsic properties. Thus the ontology must permit many many sub headings in some cases one for each segment.
Segmental differences in other repeating motor neurons
I [RC] believe that all motor neuron in the leech are in fact like the heart motor neurons. For example it is unlikely that the DE3(15) and the DE3(3) motor neuron receive the same mix of presynaptic inputs. To say they all receive inputs from some interneuron is a gross simplification of the information we want to store in neuron bank; the first may receive input from the interneuron +-6 ganglia away, while the other will certainly not because the interneuron may not exist anteriorly of the second free segmental ganglion (M2).
It is best to have a sub heading for each segment for all neurons in my opinion.
I [RC] am very picky about nomenclature of neurons. I think that all neurons should have a flexible descriptive name that reflects the current state of knowledge about them and that is used in the literature and that using acronyms in the literature just drives away readers so the flexible name is preferred in a paper. Thus what you call "HN" neurons are called "heart interneurons" in my papers and the acronym is used only to label figures or when indexing is needed. (e.g. "heart interneuron", or "coordinating interneurons" - heart implied by context, or, for indexing, the "HN(L,6) heart interneuron"). I would like to have the ontology explicitly be able to accommodate such nomenclature. Thus there is an acronymic name DE3(L,15) but the real name is "dorsal excitatory longitudinal motor neuron #3". This latter name should be flexible so that when a cells properties are expanded the name can evolve. For example, right now the HN(15) and HN(16) heart interneurons might be called "heart interneurons", but when we learn more about them we may want to call them "premotor heart interneurons" and if this occurs then we will want to change the flexible name of the HN(6) and HN(7) heart interneurons from "rear premotor heart interneurons" to simply "premotor heart interneurons" of maybe "middle premotor heart interneurons".
Note by DW
I like this idea a lot. For the database, the cell might have "HN(15)" as an AccessionID, and "heart interneuron" as DescriptiveName. Then, the DescriptiveName can be changed without affecting the connections within the database, which would be based on AccessionID.
As an aside: RC, following John Nicholls prefers that acronyms are used only as adjectives, never as nouns. Call them "P cell", not just "P", and non-leech people will find it easier to read our papers. DW thinks this is very relevant for the Wiki, but less so for the Database, because it is clear from context that all the entities here are cells.
We must also devise an agreement about how we index. I [RC] have been a trouble make here because I have resisted changing to things like "HN(L,m3) interneuron". When you do the Rz cells you will come up against this right away because they span the free segmental ganglia (m) and the brain and have different properties across the divide.
Also for nomenclature we must decide whether to use motor neuron or motoneurons: I greatly prefer the motor neuron option and believe it is widely used.
Leech Neuronal Ontology Meeting June 26-27, 2008 at the Marine Biological Lab in Woods Hole MA.