xpliu16
commented
1 year ago Hi Nicolo,
Regarding 1. Unfortunately, get_target_loc_and_dist is just pseudocode for the user defined custom connectivity rule that is passed to add_properties to generate the synapse locations. Here, we are trying to illustrate that you can assign multiple properties simultaneously.
Regarding 2. `This is the correct behavior, but it is confusing. The implied nsyns after add_properties is 1, which is the default if no nsyns is specified. Since add_properties gives a unique property to each edge, each edge is its own entry in the table. So for instance, if previously we had 1 row representing synapses from neuron A to neuron B with nsyns 5, after add_properties, there would instead be 5 lines representing each synapse, with their individual locations.
In terms of being able to help with your overall goal, do you want to see the locations as relative to the neuron or in absolute space coordinates?
Thanks, Ping
nicolomeneghetti
Dear all,
I was trying to compute the distribution of synapses' location along the z-axis onto a family of post-synaptic neurons.
As far as I've understood, to do so, you are supposed to work with the 'add_properties' of the NetworkBuilder by doing something like this:
`conns = net.add_edges( source={'type': 'basket'}, target={'type': 'pyramidal', 'model_type': 'biophysical'}, connection_rule=16, syn_weight=0.00025, delay=2.0, dynamics_params='GABA_InhToExc.json', model_template='Exp2Syn', )
conns.add_properties( ['afferent_section_id', 'afferent_section_pos', 'afferent_swc_id', 'afferent_swc_pos'], rule=rand_syn_locations, rule_params={ 'sections': ['basal', 'apical', 'axon'], 'distance_range': [30.0, 100.0], 'morphology_dir': 'components/morphologies' }, dtypes=[np.int, np.float, np.int, np.float] )`
Then you can work with the _edges.h5 file and the .swc morphology file to get the synapses’ location along the desired axis. So far (I guess) so good.
1) The first question is whether there is a simpler workaround to get the location of the synapses. In the bmtk.builder package page there is something that seems to resemble what I was looking for. Unfortunately the information there seems to be missing:
From: https://alleninstitute.github.io/bmtk/bmtk/bmtk.builder.html#:~:text=.float)-,We,-can%20also%20add `def get_target_loc_and_dist(source, target): ... return trg_loc, trg_dist
cm.add_properties(['syn_location', 'syn_distance'], rule=get_target_loc_and_dist, dtypes=[string, np.float])`
As you can see the get_target_loc_and_dist function is left empty.
2) The second question is more worrisome for me. When I run the add_edges of the bmtk builder without the add_properties:
net.add_edges( source={'type': 'basket'}, target={'type': 'pyramidal', 'model_type': 'biophysical'}, connection_rule=16, target_sections=['basal', 'apical', 'axon'], distance_range=[30.0, 100.0], .... )I get into the resulting _edges.h5 file a key field called 'nsyns'. As far as I have understood the 'nsyns' contains the number of connections from source to target for each edge (am I wrong?). Unfortunately, however, if I run the code from above (the one with the 'add_properties') the 'nsyns' disappear. So my question is: can I get both the information? Specifically, am I able to get both the location of each edge and its numerosity (that if I stand correct is contained within the 'nsyns' field)?
Thanks in advance for any help.
Best regards,
Nicolò