1. aisynphys
  2. doc

Measuring Synaptic Kinetics

In this notebook we use the Allen Institute synaptic physiology dataset to measure the kinetic properties of synaptic connections and the relationship to cell subclass.

For an introduction to the Jupyter Notebook interface interface, try Codeacademy: How To Use Jupyter Notebooks or Jupyter Notebook Quick Start Guide.

In [1]:
import numpy as np
from aisynphys.database import SynphysDatabase
from aisynphys.cell_class import CellClass
In [2]:
# Download and cache the sqlite file for the requested database
#   (for available versions, see SynphysDatabase.list_versions)
db = SynphysDatabase.load_current('small')

We are going to compare the strength of excitatory connectivity onto the three inhibitory cell subclassess -- Pvalb, Sst, and Vip.

Begin by defining these subclasses:

In [3]:
post_classes = {
    'pvalb': CellClass(cre_type='pvalb'),
    'sst': CellClass(cre_type='sst'),
    'vip': CellClass(cre_type='vip'),
}

Query the database to get all excitatory synapses with a specific postsynaptic cre type. We also filter here for "standard multipatch" experiment types in order to exclude other experiment types as well as filter for "mouse" as the species.

In [4]:
# query once for each postsynaptic type, building up a Pandas dataframe

pairs = None
for name, post_class in post_classes.items():
    pair_query = db.pair_query(
        experiment_type='standard multipatch',
        species='mouse',
        post_class=post_class,
        synapse=True,
        synapse_type='ex',
    )
    pair_query = pair_query.add_columns(
        db.Synapse.latency,
        db.Synapse.psc_rise_time,
        db.Synapse.psc_decay_tau,
        db.Synapse.psp_amplitude,
    )
    df = pair_query.dataframe()
    df['post_class'] = name
    if pairs is None:
        pairs = df
    else:
        pairs = pairs.append(df)
    print("%s: %d synapses" % (name, len(df)))

pvalb: 144 synapses
sst: 118 synapses
vip: 30 synapses

/home/luke/docs/aisynphys/aisynphys/database/database.py:758: SAWarning: TypeDecorator FloatType() will not produce a cache key because the ``cache_ok`` flag is not set to True.  Set this flag to True if this type object's state is safe to use in a cache key, or False to disable this warning.
  recs = self.all()
In [5]:
# we now have all synapses loaded into one dataframe:
pairs.head()
Out[5]:
pair.id pair.experiment_id pair.pre_cell_id pair.post_cell_id pair.has_synapse pair.has_polysynapse pair.has_electrical pair.crosstalk_artifact pair.n_ex_test_spikes pair.n_in_test_spikes pair.distance pair.lateral_distance pair.vertical_distance pair.reciprocal_id pair.meta synapse.latency synapse.psc_rise_time synapse.psc_decay_tau synapse.psp_amplitude post_class
0 16618 355 2534 2524 True False False NaN 60 60 0.000071 NaN NaN 16574 None NaN NaN NaN NaN pvalb
1 13143 274 1981 1978 True False False NaN 936 720 0.000076 0.000004 0.000019 13093 None 0.001243 0.001220 0.002434 0.000540 pvalb
2 58864 1723 10233 10220 True False False NaN 384 72 0.000049 0.000004 0.000038 58700 None 0.000862 NaN NaN 0.001326 pvalb
3 13103 274 1979 1978 True False False NaN 531 443 0.000067 0.000049 0.000011 13090 None 0.001322 0.000901 0.004337 NaN pvalb
4 22127 480 3384 3378 True False False NaN 690 690 0.000139 0.000090 0.000072 22011 None 0.000836 NaN NaN 0.000122 pvalb
In [6]:
import matplotlib.pyplot as plt
import seaborn as sns
%matplotlib inline

limits = {
    'latency': (0.3e-3, 5e-3),
    'psc_rise_time': (0.3e-3, 5e-3),
    'psc_decay_tau': (0.3e-3, 15e-3),
    'psp_amplitude': (10e-6, 5e-3),
}
units = ['ms', 'ms', 'ms', 'mV']
ticks = {
    'ms': 0.5e-3 * 2**np.arange(4),
    'mV': 20e-6 * 4**np.arange(4),
}

fig,ax = plt.subplots(len(limits), 1, figsize=(8, 2.5 * len(limits)))


for i, (field, lim) in enumerate(limits.items()):
    # box and swarm plots
    ax[i].set_yscale('log')
    ax[i].set_ylim(*lim)
    sns.boxplot(x="post_class", y="synapse."+field, data=pairs, ax=ax[i], palette='pastel', width=0.8, linewidth=1)
    sns.swarmplot(x="post_class", y="synapse."+field, data=pairs, ax=ax[i], size=5, color=(0, 0, 0, 0.5))
    
    # clean up labels / ticks
    if i == len(limits)-1:
        ax[i].set_xlabel('postsynaptic class', size=14)
    else:
        ax[i].set_xlabel('')
    ax[i].set_ylabel(field.replace('_', ' ') + ' (%s)'%units[i], size=12)
    ax[i].set_yticks(ticks[units[i]])
    ax[i].set_yticklabels(['%0.2f' % (x*1000) for x in ticks[units[i]]])
    ax[i].set_yticklabels([], minor=True)
In [7]:
fig,ax = plt.subplots()

x_field = 'psc_rise_time'
y_field = 'psc_decay_tau'
ax.set_xlim(limits[x_field])
ax.set_ylim(limits[y_field])
ax.set_yscale('log')
ax.set_xscale('log')

sns.scatterplot(
    x="synapse."+x_field, 
    y="synapse."+y_field, 
    hue="post_class", 
    size="synapse.psp_amplitude", 
    data=pairs, 
    ax=ax,
)
ax.legend(loc="upper right", bbox_to_anchor=(1.5, 1))
Out[7]:
<matplotlib.legend.Legend at 0x7fb90a51cf40>
In [ ]: