import numpy as np
from ..strategy import ConnectionStrategy
from ...helpers import DistributionConfiguration
from ...functions import get_distances
from scipy.stats.distributions import truncexpon
[docs]class ConnectomeGlomerulusGolgi(ConnectionStrategy):
"""
Legacy implementation for the connections between Golgi cells and glomeruli.
"""
defaults = {"detailed": False, "contacts": DistributionConfiguration.cast(1)}
casts = {"detailed": bool, "contacts": DistributionConfiguration.cast}
[docs] def validate(self):
if self.detailed:
morphologies = self.to_cell_types[0].list_all_morphologies()
if not morphologies:
raise ConfigurationError(
"Can't create detailed glomerulus to Golgi connections without any morphologies for the Golgi cell."
)
elif len(morphologies) > 1:
raise NotImplementedError(
"Detailed glomerulus to Golgi connections can only be made for a single morphology."
+ " (Requires the selection of morphologies to be moved from the connection module to the placement module)"
)
mr = self.scaffold.morphology_repository
morphology = mr.get_morphology(morphologies[0])
self.dendritic_compartments = morphology.get_compartments(["dendrites"])
self.morphology = morphology
def connect(self):
# Gather information for the legacy code block below.
glomerulus_cell_type = self.from_cell_types[0]
golgi_cell_type = self.to_cell_types[0]
glomeruli = self.scaffold.cells_by_type[glomerulus_cell_type.name]
golgis = self.scaffold.cells_by_type[golgi_cell_type.name]
first_glomerulus = int(glomeruli[0, 0])
r_goc_vol = golgi_cell_type.morphology.dendrite_radius
if self.detailed:
compartments = np.ones((0, 2)) * -1
comps = self.dendritic_compartments
total_compartments = len(comps)
def connectome_glom_goc(first_glomerulus, glomeruli, golgicells, r_goc_vol):
nonlocal compartments
glom_bd = np.zeros((0, 2))
glom_x = glomeruli[:, 2]
glom_y = glomeruli[:, 3]
glom_z = glomeruli[:, 4]
# If synaptic contacts need to be made we use this exponential distribution
# to pick the closer by compartments.
exp_dist = truncexpon(b=5, scale=0.03)
# for all Golgi cells: calculate which glomeruli fall into the volume of GoC
# basolateral dendrites, then choose 40 of them for the connection and delete
# them from successive computations, since 1 axon is connected to 1 GoC
for golgi_id, golgi_type, golgi_x, golgi_y, golgi_z in golgicells:
# Geometric constraints: glom less than `r_goc_vol` away from golgi and
# golgi cell soma above glom.
volume_matrix = (
((glom_x - golgi_x) ** 2)
+ ((glom_y - golgi_y) ** 2)
+ ((glom_z - golgi_z) ** 2)
- (r_goc_vol ** 2)
).__le__(0) & glom_y.__le__(golgi_y)
# finds indexes of granules that can potentially be connected
good_gloms = np.where(volume_matrix == True)[0]
# Translate local id to simulation id
connected_gloms = good_gloms + first_glomerulus
if self.detailed:
# Draw a sample from the configured contacts distribution for each
# connected glomerulus.
samples = [int(x) for x in self.contacts.draw(len(connected_gloms))]
# The total synaptic contacts is the sum of the contacts with each
# glomerulus.
total_contacts = sum(samples)
# Compose an empty connection matrix to hold the connection records
matrix = np.zeros((total_contacts, 2))
# Holds the connection_compartments for this golgi cell
gg_comps = np.ones(matrix.shape) * -1
# Draw rolls from the exponential distribution equal to the total amount
# of synaptic contacts to be made between this Golgi cell and all its
# glomeruli.
rolls = exp_dist.rvs(size=total_contacts)
pointer = 0
for i, cg in enumerate(connected_gloms):
sample = samples[i]
_end = pointer + sample
# Fill in `sample` records of the connection matrix
matrix[pointer:_end, 0] = cg
matrix[pointer:_end, 1] = golgi_id
gid = good_gloms[i]
# Get the distance sorted compartment indices
d = np.array(
get_distances(
[c.start for c in comps],
[
glom_x[gid] - golgi_x,
glom_y[gid] - golgi_y,
glom_z[gid] - golgi_z,
],
)
)
d_comps = np.argsort(d)
# Pick compartments according to a exponential distribution mapped
# through the distance indices: high chance to pick closeby comps.
gg_comps[pointer:_end, 1] = [
comps[d_comps[int(k * total_compartments)]].id
for k in rolls[pointer:_end]
]
pointer = _end
# Stack the connection matrix
glom_bd = np.vstack((glom_bd, matrix))
# Stack the compartment matrix
compartments = np.vstack((compartments, gg_comps))
else:
matrix = np.zeros((len(connected_gloms), 2))
matrix[:, 0] = connected_gloms # from cell
matrix[:, 1] = golgi_id # to cell
glom_bd = np.vstack((glom_bd, matrix))
return glom_bd
connectome = connectome_glom_goc(first_glomerulus, glomeruli, golgis, r_goc_vol)
if self.detailed:
morphologies = np.zeros(connectome.shape)
morphology_map = [self.morphology.morphology_name]
self.scaffold.connect_cells(
self,
connectome,
compartments=compartments,
morphologies=morphologies,
morpho_map=morphology_map,
)
else:
self.scaffold.connect_cells(self, connectome)
