def plot_cell_summary(rec1, rec2):
"""
Plot one page summary of single cell responses for BVT tone discrimination task.
rec1 is high sampling rate (for raster plotting)
rec2 is low sampling rate (for other plots, such as psth's)
Produce a one-page summary of each cell. This includes:
1) Time courses:
pupil trace
target responses (smoothed)
refs responses nearest tar freqs?
2) FTC per file
w/ and w/o pupil regression
3) PSTH / raster per file
w/ and w/o pupil regression
"""
# apply mask to recording and get global params, like tar list
rec1 = rec1.apply_mask(reset_epochs=True)
rec2 = rec2.apply_mask(reset_epochs=True)
targets = [str(t) for t in np.sort([t for t in rec1.epochs.name.unique() if 'TAR_' in t])]
# layout figure grid (define axes objects)
files = [f for f in rec1.epochs.name.unique() if 'FILE_' in f]
nf = len(files)
fig = plt.figure(figsize=(16, 12))
pupil_trace = plt.subplot2grid((6, nf), (0, 0), colspan=nf)
target_resp = plt.subplot2grid((6, nf), (1, 0), colspan=nf)
ftc_pl = []
psth = []
ftc_pl_pr = []
psth_pr = []
for i in range(nf):
ftc_pl.append(plt.subplot2grid((6, nf), (2, i), colspan=1))
psth.append(plt.subplot2grid((6, nf), (3, i), colspan=1))
ftc_pl_pr.append(plt.subplot2grid((6, nf), (4, i), colspan=1))
psth_pr.append(plt.subplot2grid((6, nf), (5, i), colspan=1))
# plot time courses
time = np.arange(0, rec2['pupil'].shape[-1] / rec2['resp'].fs, 1 / rec2['resp'].fs)
# pupil time course
pupil_trace.plot(time, rec2['pupil']._data.T, color='green', label='pupil')
act = rec2['resp'].epoch_to_signal('ACTIVE_EXPERIMENT')._data.squeeze() * rec2['pupil']._data.max()
pupil_trace.fill_between(time, act, color='lightgrey', alpha=0.5)
pupil_trace.legend(fontsize=8, frameon=False)
pupil_trace.set_xlabel('Time (s)')
pupil_trace.set_ylabel('Size (pixels)')
# target resp time course
for t in targets:
ti, resp = get_resp_timecourse(rec2, str(t))
resp_roll = rolling_window(resp, 5)
ti_roll = rolling_window(ti, 5)
resp = np.mean(resp_roll, axis=-1)
ti = np.median(ti_roll, axis=-1)
resp_sem = np.std(resp_roll, axis=-1) / np.sqrt(resp_roll.shape[-1])
target_resp.plot(ti, resp, label=t)
target_resp.fill_between(ti, resp-resp_sem, resp+resp_sem,
color=target_resp.get_lines()[-1].get_color(),
alpha=0.3, lw=0)
lim = target_resp.get_ylim()[-1]
act = rec2['resp'].epoch_to_signal('ACTIVE_EXPERIMENT')._data.squeeze() * lim
target_resp.fill_between(time, act, color='lightgrey', alpha=0.5)
target_resp.legend(fontsize=8, frameon=False)
target_resp.set_xlabel('Time (s)')
target_resp.set_ylabel('Spk / bin')
# pupil regressed recording
pr = preproc.regress_state(rec2, state_sigs=['pupil'], regress=['pupil'])
ylim = cplt.get_ylim(rec1, fs=20, epochs=targets)
tr_max = cplt.get_tr_max(rec1, epochs=targets)
ftc_ylim = 0
ftc_pr_ylim = 0
for i, f in enumerate(files):
r1 = rec1.copy()
r1 = r1.and_mask([f]).apply_mask(reset_epochs=True)
r2 = rec2.copy()
r2 = r2.and_mask([f]).apply_mask(reset_epochs=True)
r2_pr = pr.copy()
r2_pr = r2_pr.and_mask([f]).apply_mask(reset_epochs=True)
# plot ftc
ftc = tc.get_tuning_curves(r2)
vals = ftc.loc['r'].to_numpy(dtype=np.float).squeeze()
sem = ftc.loc['sem'].to_numpy(dtype=np.float).squeeze()
freqs = ftc.loc['r'].columns
ftc_pl[i].errorbar(freqs, vals, sem, color='k')
for tar in targets:
tar = int(tar.split('_')[-1])
ftc_pl[i].axvline(tar, color='r', linestyle='--')
if ftc_pl[i].get_ylim()[-1] > ftc_ylim:
ftc_ylim = ftc_pl[i].get_ylim()[-1]
ftc_pl[i].set_ylabel('Spk / bin')
ftc_pl[i].set_xlabel("CF")
# plot psth
cplt.plot_raster_psth(r1, targets, psth_fs=20, ax=psth[i], ylim=ylim, raster=True, tr_max=tr_max)
# plot ftc pupil regressed
ftc = tc.get_tuning_curves(r2_pr)
vals = ftc.loc['r'].to_numpy(dtype=np.float).squeeze()
sem = ftc.loc['sem'].to_numpy(dtype=np.float).squeeze()
freqs = ftc.loc['r'].columns
ftc_pl_pr[i].errorbar(freqs, vals, sem, color='k')
for tar in targets:
tar = int(tar.split('_')[-1])
ftc_pl_pr[i].axvline(tar, color='r', linestyle='--')
if ftc_pl_pr[i].get_ylim()[-1] > ftc_pr_ylim:
ftc_pr_ylim = ftc_pl_pr[i].get_ylim()[-1]
ftc_pl_pr[i].set_ylabel('Spk / bin')
ftc_pl_pr[i].set_xlabel("CF")
# plot psth pupil regressed
cplt.plot_raster_psth(r2_pr, targets, psth_fs=20, ax=psth_pr[i], ylim=ylim, raster=False)
# set ylim for FTC
for i in range(len(files)):
ftc_pl[i].set_ylim((0, ftc_ylim))
ftc_pl_pr[i].set_ylim((0, ftc_pr_ylim))
fig.tight_layout()
return fig
df = pd.DataFrame()
for batch in batches:
sites = np.unique([c[:7] for c in nd.get_batch_cells(batch).cellid])
sites = [s for s in sites if s != 'CRD013b']
for site in sites:
manager = BAPHYExperiment(batch=batch, siteid=site)
rec = manager.get_recording(recache=recache, **options)
rec['resp'] = rec['resp'].rasterize()
# mask appropriate trials
rec = rec.and_mask(
['PASSIVE_EXPERIMENT', 'HIT_TRIAL', 'CORRECT_REJECT_TRIAL'])
rec = rec.apply_mask(reset_epochs=True)
if regress_pupil:
rec = preproc.regress_state(rec, state_sigs=['pupil'])
ra = rec.copy()
ra = ra.create_mask(True)
ra = ra.and_mask(['HIT_TRIAL', 'CORRECT_REJECT_TRIAL'])
rp = rec.copy()
rp = rp.create_mask(True)
rp = rp.and_mask(['PASSIVE_EXPERIMENT'])
# stim pairs
targets = thelp.sort_targets(
[f for f in ra['resp'].epochs.name.unique() if 'TAR_' in f])
targets = [
t for t in targets if (ra.apply_mask(
reset_epochs=True)['resp'].epochs.name == t).sum() >= 5
rec = manager.get_recording(recache=recache, **options)
rec['resp'] = rec['resp'].rasterize()
if batch == 302:
c, _ = parse_cellid({'cellid': site, 'batch': batch})
rec['resp'] = rec['resp'].extract_channels(c)
behavior_performance = manager.get_behavior_performance(**options)
allop = copy.deepcopy(options)
allop['keep_following_incorrect_trial'] = True
allop['keep_cue_trials'] = True
allop['keep_early_trials'] = True
behavior_performance_all = manager.get_behavior_performance(**allop)
# regress out first order pupil
if regress_pupil & regress_task:
rec = preproc.regress_state(rec, state_sigs=['pupil', 'behavior'])
elif regress_pupil:
rec = preproc.regress_state(rec, state_sigs=['pupil'])
elif regress_task:
rec = preproc.regress_state(rec, state_sigs=['behavior'])
if deflate:
# brute force remove all information on delta noise correlation axis from the response
resp = rec['resp']._data
lv = pickle.load(open(DIR + '/results/drsc_axes.pickle', "rb"))
def_axis = lv[site]['tarOnly']['evecs'][:, [0]]
projection = (resp.T.dot(def_axis) @ def_axis.T).T
resp = resp - projection
rec['resp'] = rec['resp']._modified_copy(resp)
ra = rec.copy()
# only necessary if using correction method 1
if lv_regress:
if rec['lv']._data.shape[0] > 1:
rec['lv'] = rec['pupil']._modified_copy(rec['lv']._data[1:, :])
rec = rec.create_mask(True)
# filtering / pupil regression must always go first!
if pupil_regressed & lv_regress:
if regression_method1:
log.info(
'Regress first and second order pupil using brute force method'
)
rec = preproc.regress_state(rec,
state_sigs=['pupil', 'lv'],
regress=['pupil', 'lv'])
elif regression_method2:
log.info(
'Regress first and second order pupil by subtracting model pred'
)
mod_data = rec['resp']._data - rec['pred']._data + rec[
'psth_sp']._data
rec['resp'] = rec['resp']._modified_copy(mod_data)
else:
raise ValueError("No regression method specified!")
elif pupil_regressed:
if regression_method1:
log.info('Regress first order pupil using brute force method')
