Thank you for the clarification Stephanie.
I have one more question:
Considering that the experiments performed at 50 Hz could have different times between the 8th stimulus and the 9th stimulus (recovery train of 4 pulses), how can I get this time value?
Thanks
The easiest place to get this is in the MultiPatchProbe table of the database. There is a column called ârecovery_delayâ which specifies the amount of time between the 8th and 9th pulse of any stimulus frequency. Our documentation shows what is available in each table of the database and how they can be linked to one another. If you have the recording object that you are interested in you can get to multi_patch_probe via recording.patch_clamp_recording.multi_patch_probe.recovery_delay
Dear Stephanie,
I want to create a csv file with the data that I want from your database.
I would like to have: âpairâ, âsyn_typeâ, âpre_exp_idâ, âpre_idâ, âpre_classâ, âpre_layerâ,âpost_exp_idâ, âpost_idâ,
âpost_classâ, âpost_layerâ, âPSP_mean (mV)â, âstim_freq (Hz)â, âtime_rec (ms)â
Here I send you my code:
import numpy as np
import scipy.stats
from aisynphys.database import default_db as db
from aisynphys.cell_class import CellClass, classify_cells, classify_pairs
from aisynphys.connectivity import measure_connectivity, pair_was_probed
from aisynphys.database import SynphysDatabase
import matplotlib.pyplot as plt
from aisynphys.ui.notebook import plot_stim_sorted_pulse_amp
from aisynphys.dynamics import stim_sorted_pulse_amp
import csv
# Download and cache the sqlite file for the requested database
# Load all synapses associated with human projects
# First load the data base
path_to_db = '/Users/natalibarros/ai_synphys_cache/database/synphys_r2.0_medium.sqlite'
db = SynphysDatabase.load_version(path_to_db)
pairs = db.pair_query(project_name=db.human_projects, synapse=True).all()
print("loaded %d synapses" % len(pairs))
print("number of experiments %d" % len(set([pair.experiment for pair in pairs])))
print("number of experiments with recordings %d" %len(set([pair.experiment.data for pair in pairs])))
# We can do a pair_query to return pairs with stp valid values
query = db.pair_query(
experiment_type='standard multipatch', # filter: just multipatch experiments
species='human', # filter: only human data
synapse=True, # filter: only cell pairs connected by synapse
)
pairs = query.all()
print('number of pairs with patch experiments', len(pairs))
pairs_with_stp = [p for p in pairs if p.dynamics.stp_induction_50hz is not None] # Is that ok?
print('number of pairs with STP dynamics', len(pairs_with_stp))
N = 1
# create a csv file object to save useful values
path_to_csv_file = '/Users/natalibarros/Desktop/EPFL_BBP/HUMAN_CORTEX/CIRCUIT_BUILDING/DATA/DATA_Allen_Brain/PSP_values_DBmedium_04.csv'
with open(path_to_csv_file, 'wt') as csvfile:
writer = csv.writer(csvfile, delimiter=',', lineterminator='\n', )
# Add the header row
writer.writerow(['pair', 'syn_type', 'pre_exp_id', 'pre_id', 'pre_class', 'pre_layer','post_exp_id', 'post_id',
'post_class', 'post_layer', 'PSP_mean (mV)', 'stim_freq (Hz)', 'time_rec (ms)'])
for pair in pairs_with_stp:
print('#', N)
print('working on pair', pair)
N = N+1
# syn and cells information
try:
syn_type = pair.synapse.synapse_type
pre_exp_id = pair.pre_cell
pre_id = pair.pre_cell_id
pre_class = pair.pre_cell.cell_class
pre_layer = pair.pre_cell.cortical_location.cortical_layer
post_exp_id = pair.post_cell
post_id = pair.post_cell_id
post_class = pair.post_cell.cell_class
post_layer = pair.post_cell.cortical_location.cortical_layer
except AttributeError:
print('missing attribute')
continue
# PSPs information
qc_pass_data = stim_sorted_pulse_amp(pair, db=db)
# Stim freq information (ind_f)
try:
ind_f = int(qc_pass_data['induction_frequency'][0])
# scatter plots of event amplitudes sorted by pulse number
mask = qc_pass_data['induction_frequency'] == ind_f
except KeyError:
ind_f = 50 # assumption: if there is no stim freq, we suppose is 50 Hz
mask = qc_pass_data['induction_frequency'] == ind_f
print('no ind_f vector')
filtered = qc_pass_data[mask].copy()
sign = 1 if pair.synapse.synapse_type == 'ex' else -1
try:
filtered['dec_fit_reconv_amp'] *= sign * 1000 # the data are in Volts, so I convert them into mV.
except KeyError:
print('No fit amps for pair: %s' % pair)
PSP_means = filtered.groupby('pulse_number').mean()['dec_fit_reconv_amp'].to_list()
# Find recovery_delay???????
# Add the data row
writer.writerow([pair, syn_type, pre_exp_id, pre_id, pre_class, pre_layer,
post_exp_id, post_id, post_class, post_layer, PSP_means, ind_f, recov_time])
My questions are:
Could you help me please?
Hi Natali,
try: except where you are masking qc_pass_data your assumption that if the KeyError is caught that ind_f == 50 also seems incorrectqc_pass_data dataframe that you have as a column (named recovery_delay) in the dataframeqc_pass_data.head()
fit_amp dec_fit_reconv_amp baseline_dec_fit_reconv_amp qc_pass \
0 0.000136 0.000234 -0.000004 True
1 0.000508 0.000492 0.000032 True
2 0.000019 -0.000011 -0.000015 True
3 0.001213 0.000295 -0.000018 True
4 0.000276 0.000073 -0.000010 True
pulse_number induction_frequency recovery_delay sync_rec_ext_id \
0 1 50.0 0.128 41
1 2 50.0 0.128 41
2 3 50.0 0.128 41
3 4 50.0 0.128 41
4 5 50.0 0.128 41
id
0 14331746
1 14331747
2 14331749
3 14331751
4 14331753
Nominally our recovery delays are (in seconds) 0.125, 0.25, 0.5, 1.0, 4.0 but due to acquisition sampling rate the values that are populated in the Database are a little different. You can see what recovery delays are present if you do:
qc_pass_data['recovery_delay'].unique()
>>> array([0.128, 0.253, 0.503, 1.003, 4.003]
Note that itâs only the 50 Hz stimulus that has multiple recovery delays, all other frequencies are only the 0.25s recovery delay. So if you wanted to filter your qc_pass_data to only include a particular recovery you could do something like:
mask = (qc_pass_data['induction_frequency'] == 50) & (qc_pass_data['recovery_delay'] == 0.503)
filtered = qc_pass_data[mask]
filtered.head()
fit_amp dec_fit_reconv_amp baseline_dec_fit_reconv_amp qc_pass \
24 6.519115e-05 0.000028 -0.000036 True
25 1.434500e-03 0.000820 -0.000021 True
26 1.217622e-03 0.000187 0.000020 True
27 2.364060e-04 -0.000149 0.000027 True
28 3.436510e-12 -0.000027 0.000014 True
pulse_number induction_frequency recovery_delay sync_rec_ext_id \
24 1 50.0 0.503 43
25 2 50.0 0.503 43
26 3 50.0 0.503 43
27 4 50.0 0.503 43
28 5 50.0 0.503 43
id
24 14333817
25 14333819
26 14333821
27 14333823
28 14333825
Thank you very much for the answer.
I think I got it now, mean PSP amps separated by frequencies and with recovery_delay (important for 50Hz)
After the line:
# PSPs information
qc_pass_data = stim_sorted_pulse_amp(pair, db=db)
I included this:
# stim freq and recovery delay information
ind_f = qc_pass_data['induction_frequency'].unique()
rec_del = qc_pass_data['recovery_delay'].unique()
ind_f_list = ind_f.tolist()
rec_del_list = rec_del.tolist()
for i, r in zip(ind_f_list,rec_del_list):
mask = (qc_pass_data['induction_frequency'] == i) & (qc_pass_data['recovery_delay'] == r)
filtered = qc_pass_data[mask]
# detect ex or ihb and flip trace if inb
sign = 1 if pair.synapse.synapse_type == 'ex' else -1
try:
filtered['dec_fit_reconv_amp'] *= sign * 1000 # convert values into mV
except KeyError:
print('No fit amps for pair: %s' % pair)
PSP_means = filtered.groupby('pulse_number').mean()['dec_fit_reconv_amp'].to_list()
# Add the data row
writer.writerow([pair, syn_type, pre_exp_id, pre_id, pre_class, pre_layer,
post_exp_id, post_id, post_class, post_layer, PSP_means, i, r])
The zip(ind_f_list, rec_del_list) assumes that the induction frequency and recovery delay are paired (thatâs what zip() does) which isnât the case. If you want all combinations of induction frequencies and recovery delays I would do a nested loop, but since the only induction frequency that has multiple recovery delays is the 50Hz stimulus you could just create a nested if loop that goes through the recovery delays when the induction frequency is 50Hz. Based on exactly what you want there are multiple ways you could construct this but youâve at least got the frequencies and delays pulled out correctly.
Thank you Stepahnie for the clarification,
I changed the code to this.
# stim freq and recovery delay information
ind_f = qc_pass_data['induction_frequency'].unique()
#rec_del = qc_pass_data['recovery_delay'].unique()
ind_f_list = ind_f.tolist()
#rec_del_list = rec_del.tolist()
for i in ind_f_list:
if i == 50:
print('50 Hz stimulation')
rec_del = qc_pass_data['recovery_delay'].unique()
rec_del_list = rec_del.tolist()
for r in rec_del_list:
mask = (qc_pass_data['induction_frequency'] == i) & (qc_pass_data['recovery_delay'] == r)
filtered = qc_pass_data[mask]
print(r)
sign = 1 if pair.synapse.synapse_type == 'ex' else -1
try:
filtered['dec_fit_reconv_amp'] *= sign * 1000
except KeyError:
print('No fit amps for pair: %s' % pair)
PSP_means = filtered.groupby('pulse_number').mean()['dec_fit_reconv_amp'].to_list()
if PSP_means == []:
print('PSP empty')
pass
else:
print('PSP')
# Add the data row
writer.writerow([pair, syn_type, pre_exp_id, pre_id, pre_class, pre_layer,
post_exp_id, post_id, post_class, post_layer, PSP_means, i, r])
else:
print('other freq')
mask = (qc_pass_data['induction_frequency'] == i)
filtered = qc_pass_data[mask]
sign = 1 if pair.synapse.synapse_type == 'ex' else -1
try:
filtered['dec_fit_reconv_amp'] *= sign * 1000
except KeyError:
print('No fit amps for pair: %s' % pair)
PSP_means = filtered.groupby('pulse_number').mean()['dec_fit_reconv_amp'].to_list()
if PSP_means == []:
print('PSP empty')
pass
else:
print('PSP')
writer.writerow([pair, syn_type, pre_exp_id, pre_id, pre_class, pre_layer,
post_exp_id, post_id, post_class, post_layer, PSP_means, i, r])
Now instead of having around 200 pair recordings with PSP values I have 1248!!!
Does this number sound correct?
That doesnât sound unreasonable. Youâre writing enough info to your file that you could check for duplicates for the same pair. I would double check your r variable for the âother freqâ loop. Youâre writing it to your file but you arenât actually tracking it there.
To plot all synaptic traces, youâll need to iterate through your recordings and pairs. Hereâs a simplified example:
import matplotlib.pyplot as plt
from aisynphys.ui.notebook import plot_stim_sorted_pulse_amp
for recording_file in recordings:
pair = db.experiment_from_ext_id(â#exp_idâ).pairs[(â#preâ, â#postâ)]
fig, ax = plt.subplots()
plot_stim_sorted_pulse_amp(pair, ax)
plt.show()
The PyNWB error ValueError: No data_type found for builder root indicates that PyNWB is having trouble reading the NWB file correctly for CRM enrichment. Ensure that you are using the correct PyNWB version that is compatible with your NWB file version.
Hi Natali,
I am working with the synaptic physiology dataset as well, and I am running into this exact error where I cannot import aisynphys because there is no module neuroanalysis.
I tried running pip install neuroanalysis, and it seems I have the module, but now I run into the error that:
TypeError: 'type' object is not subscriptable
Do you by chance know how to best install neuroanalysis?
Hello Jeff,
Itâs been a while since I donât use this, but I donât remember having any TypeError.
I basically followed the information in here: Installation â aisynphys documentation
and everything worked fine for me regarding the installation.
Hi Natali!
Thank you for your help!
Kind regards,
Jeffrey
If you keep seeing this error, please post the complete tracback and a little bit about what you were doing that led to the error.
Luke