search

open-connectome-classes / StatConn-Spring-2015-Info

introductory material
18 stars 4 forks source link

Axo-axonic synapses in terms of graph theory #4

Open whock opened 9 years ago

whock commented 9 years ago

Axo-axonic synapses are connections between the axon of one neuron and the axon of another neuron, unlike 'typical' synapses which are connections between an axon and a dendrite (or cell body). How would this be interpreted in terms of graph theory?

Link to a pic with the different kinds of synapses: http://bit.ly/1tuNwCo

jovo commented 9 years ago

great question! what do you think? consider directed vs. undirected, etc...

On Wed Jan 28 2015 at 9:55:54 PM whock notifications@github.com wrote:

Axo-axonic synapses are connections between the axon of one neuron and the axon of another neuron, unlike 'typical' synapses which are connections between an axon and a dendrite (or cell body). How would this be interpreted in terms of graph theory?

Link to a pic with the different kinds of synapses: http://bit.ly/1tuNwCo

— Reply to this email directly or view it on GitHub https://github.com/Statistical-Connectomics-Sp15/intro/issues/4.

whock commented 9 years ago

Not sure yet, but I think it may have to do with whether we let axo-axonic synapses (AAS) have an implicit or explicit representation. What I mean is that AAS usually (always?) modulate some other synapse so instead of drawing in AAS as some type of edge or something we can take them into account by the weighting of the node whose edge the AAS modifies (assuming a weighted graph).

If we want them to be accounted for explicitly (i.e. actually have an edge or node to describe them) we could ignore that they contact an axon ('edge') and have them just contact the cell ('node') because, again, they are mostly modulatory if I understand correctly. Otherwise we'd have to introduce different types of nodes instead of assuming all nodes are the same. There would be main nodes (cell bodies/dendrites) and then minor nodes (the nodes that come about when an edge contacts another edge - the AAS).

But at this point I don't know enough to speculate so I'll keep it in mind as we continue the course :)

edunnwe1 commented 9 years ago

I'm new to neuroscience, but from the image, it seems that only one neuron is able to synapse on the other, so I think that a directed graph would make sense here: If the nodes are neurons, then the edge between them would have some direction since one neuron influences the activity of another, but not vice-versa. But, I'm not sure if that's accurate, and I haven't found a good resource on it, so I'd be interested to know if that's actually true for axoaxonic synapses. I think the idea that you suggested of incorporating the axoaxonic synapses as a node attribute (i.e. weighting of the neuron receving the input, so perhapse the node attributes are input current) is interesting but I wonder if it could become problematic if the axoaxonic synapse only happens sometimes. In that case having the property tied in the node doesn't reflect the variability in the neuron's input. Instead maybe you could capture it as an edge attribute, and talk about the axoaxonic spike between neuron A and neuron B as a conditional probability?