In [1]:
%load_ext autoreload
%autoreload 2
In [2]:
import os
import expipe
import pathlib
import numpy as np
import spatial_maps.stats as stats
import septum_mec
import septum_mec.analysis.data_processing as dp
import septum_mec.analysis.registration
import head_direction.head as head
import spatial_maps as sp
import speed_cells.speed as spd
import re
import joblib
import multiprocessing
import shutil
import psutil
import pandas as pd
import matplotlib.pyplot as plt
import matplotlib
# import seaborn as sns
from distutils.dir_util import copy_tree
from neo import SpikeTrain
import scipy
from tqdm.notebook import tqdm_notebook as tqdm
tqdm.pandas()
from spike_statistics.core import permutation_resampling
from spikewaveform.core import calculate_waveform_features_from_template, cluster_waveform_features
from septum_mec.analysis.plotting import violinplot, despine
In [3]:
%matplotlib inline
plt.rc('axes', titlesize=12)
plt.rcParams.update({
'font.size': 12,
'figure.figsize': (6, 4),
'figure.dpi': 150
})
output_path = pathlib.Path("output") / "stimulus-response"
(output_path / "statistics").mkdir(exist_ok=True, parents=True)
(output_path / "figures").mkdir(exist_ok=True, parents=True)
output_path.mkdir(exist_ok=True)
In [4]:
data_loader = dp.Data()
actions = data_loader.actions
project = data_loader.project
In [5]:
identification_action = actions['identify-neurons']
sessions = pd.read_csv(identification_action.data_path('sessions'))
units = pd.read_csv(identification_action.data_path('units'))
session_units = pd.merge(sessions, units, on='action')
In [6]:
stim_action = actions['stimulus-response']
stim_results = pd.read_csv(stim_action.data_path('results'))
In [7]:
# lfp_results has old unit id's but correct on (action, unit_name, channel_group)
stim_results = stim_results.drop('unit_id', axis=1)
In [8]:
statistics_action = actions['calculate-statistics']
shuffling = actions['shuffling']
statistics_results = pd.read_csv(statistics_action.data_path('results'))
statistics_results = session_units.merge(statistics_results, how='left')
quantiles_95 = pd.read_csv(shuffling.data_path('quantiles_95'))
action_columns = ['action', 'channel_group', 'unit_name']
data = pd.merge(statistics_results, quantiles_95, on=action_columns, suffixes=("", "_threshold"))
In [9]:
data['unit_day'] = data.apply(lambda x: str(x.unit_idnum) + '_' + x.action.split('-')[1], axis=1)
In [10]:
data = data.merge(stim_results, how='left')
In [11]:
waveform_action = actions['waveform-analysis']
waveform_results = pd.read_csv(waveform_action.data_path('results')).drop('template', axis=1)
In [12]:
data = data.merge(waveform_results, how='left')
In [13]:
colors = ['#d95f02','#e7298a']
labels = ['11 Hz', '30 HZ']
queries = ['frequency==11', 'frequency==30']
In [14]:
data.bs = data.bs.astype(bool)
In [15]:
data.loc[data.eval('t_i_peak == t_i_peak and not bs'), 'ns_inhibited'] = True
data.ns_inhibited.fillna(False, inplace=True)
data.loc[data.eval('t_i_peak != t_i_peak and not bs'), 'ns_not_inhibited'] = True
data.ns_not_inhibited.fillna(False, inplace=True)
In [16]:
query = (
'gridness > gridness_threshold and '
'information_rate > information_rate_threshold and '
'gridness > .2 and '
'average_rate < 25'
)
sessions_above_threshold = data.query(query)
print("Number of sessions above threshold", len(sessions_above_threshold))
print("Number of animals", len(sessions_above_threshold.groupby(['entity'])))
In [17]:
gridcell_sessions = data[data.unit_day.isin(sessions_above_threshold.unit_day.values)]
print("Number of gridcells", gridcell_sessions.unit_idnum.nunique())
print("Number of gridcell recordings", len(gridcell_sessions))
print("Number of animals", len(gridcell_sessions.groupby(['entity'])))
In [18]:
data.loc[:,'gridcell'] = np.nan
data['gridcell'] = data.isin(gridcell_sessions)
data.loc[data.eval('not gridcell and bs'), 'bs_not_gridcell'] = True
data.bs_not_gridcell.fillna(False, inplace=True)
In [19]:
data.query('baseline and Hz11 and gridcell').head()
Out[19]:
In [20]:
entity_date_ = data.query('stim_location=="ms"').entity_date.unique()
In [21]:
data = data.query('stim_location!="mecl" and stim_location!="mecr"')
In [22]:
density = True
cumulative = True
histtype = 'step'
lw = 2
bins = {
't_i_peak': np.arange(0, 0.02, 0.0005),
't_e_peak': np.arange(0, 0.03, 0.0005),
'p_i_peak': np.arange(0, 0.04, 0.0005),
'p_e_peak': np.arange(0, 0.04, 0.0005),
}
xlabel = {
't_i_peak': 's',
't_e_peak': 's',
'p_i_peak': 'prob',
'p_e_peak': 'prob',
}
for cell_type in ['gridcell', 'bs_not_gridcell', 'ns_inhibited', 'ns_not_inhibited']:
for key in bins:
fig = plt.figure(figsize=(3.5,2.2))
plt.suptitle(key + ' ' + cell_type)
legend_lines = []
for color, query, label in zip(colors, queries, labels):
data.query(query + ' and ' + cell_type)[key].hist(
bins=bins[key], density=density, cumulative=cumulative, lw=lw,
histtype=histtype, color=color)
legend_lines.append(matplotlib.lines.Line2D([0], [0], color=color, lw=lw, label=label))
plt.xlabel(xlabel[key])
plt.legend(
handles=legend_lines,
bbox_to_anchor=(1.04,1), borderaxespad=0, frameon=False)
plt.tight_layout()
plt.grid(False)
plt.xlim(-0.001, bins[key].max() - bins[key].max()*0.02)
despine()
figname = f'histogram-{key}-{cell_type}'.replace(' ', '-')
fig.savefig(
output_path / 'figures' / f'{figname}.png',
bbox_inches='tight', transparent=True)
fig.savefig(
output_path / 'figures' / f'{figname}.svg',
bbox_inches='tight', transparent=True)
In [23]:
from septum_mec.analysis.plotting import plot_bootstrap_timeseries
In [24]:
psth = pd.read_feather(output_path / 'data' / 'psth.feather')
times = pd.read_feather(output_path / 'data' / 'times.feather')
In [25]:
times = times.T.iloc[0].values
In [26]:
cs = ['#993404', '#980043', '#d95f02', '#e7298a']
lb = ['GC 11 Hz', 'GC 30 Hz', 'NSI 11 Hz', 'NSI 30 Hz']
fig, axs = plt.subplots(1, 2, sharex=True, sharey=True, figsize=(5,2))
ii = 0
for cell_type, ls in zip(['gridcell', 'ns_inhibited'], ['-', '--']):
for i, (ax, query) in enumerate(zip(axs.ravel(), queries)):
selection = [
f'{r.action}_{r.channel_group}_{r.unit_name}'
for i, r in data.query(query + ' and ' + cell_type).iterrows()]
values = psth.loc[:, selection].dropna(axis=1).to_numpy()
plot_bootstrap_timeseries(times*1000, values, ax=ax, lw=2, label=lb[ii], color=cs[ii], ls=ls)
# ax.set_title(titles[i])
ax.set_xlabel('Time (ms)')
ax.legend(frameon=False)
ii += 1
axs[0].set_ylabel('Probability density')
despine()
plt.xlim(-5, 30)
figname = f'response-probability-gc-ns'
fig.savefig(
output_path / 'figures' / f'{figname}.png',
bbox_inches='tight', transparent=True)
fig.savefig(
output_path / 'figures' / f'{figname}.svg',
bbox_inches='tight', transparent=True)
In [27]:
cs = ['#d95f02', '#e7298a', '#993404', '#980043']
lb = ['NSI 11 Hz', 'NSI 30 Hz', 'NS 11 Hz', 'NS 30 Hz']
fig, axs = plt.subplots(1, 2, sharex=True, sharey=True, figsize=(5,2))
ii = 0
for cell_type, ls in zip(['ns_inhibited', 'ns_not_inhibited'], ['--', '-']):
for i, (ax, query) in enumerate(zip(axs.ravel(), queries)):
selection = [
f'{r.action}_{r.channel_group}_{r.unit_name}'
for i, r in data.query(query + ' and ' + cell_type).iterrows()]
values = psth.loc[:, selection].dropna(axis=1).to_numpy()
plot_bootstrap_timeseries(times*1000, values, ax=ax, lw=2, label=lb[ii], color=cs[ii], ls=ls)
# ax.set_title(titles[i])
ax.set_xlabel('Time (ms)')
ax.legend(frameon=False)
ii += 1
axs[0].set_ylabel('Probability density')
despine()
plt.xlim(-5, 30)
figname = f'response-probability-nsi-ns'
fig.savefig(
output_path / 'figures' / f'{figname}.png',
bbox_inches='tight', transparent=True)
fig.savefig(
output_path / 'figures' / f'{figname}.svg',
bbox_inches='tight', transparent=True)
In [ ]:
Store results in Expipe action¶
In [28]:
action = project.require_action("stimulus-response")
In [29]:
copy_tree(output_path, str(action.data_path()))
Out[29]:
In [30]:
septum_mec.analysis.registration.store_notebook(action, "20_stimulus-spike-response.ipynb")
In [ ]: