def make(self, key):
water_res_num, sess_date = get_wr_sessdate(key)
sess_dir = store_stage / water_res_num / sess_date
sess_dir.mkdir(parents=True, exist_ok=True)
# ---- Setup ----
time_period = (-0.4, 0)
probe_keys = (ephys.ProbeInsertion & key).fetch(
'KEY', order_by='insertion_number')
fig1, axs = plt.subplots(len(probe_keys),
len(probe_keys),
figsize=(16, 16))
if len(probe_keys) > 1:
[a.axis('off') for a in axs.flatten()]
# ---- Plot Coding Direction per probe ----
probe_proj = {}
for pid, probe in enumerate(probe_keys):
units = ephys.Unit & probe
label = (ephys.ProbeInsertion & probe).aggr(
ephys.ProbeInsertion.RecordableBrainRegion.proj(
brain_region='CONCAT(hemisphere, " ", brain_area)'),
brain_regions='GROUP_CONCAT(brain_region SEPARATOR", ")'
).fetch1('brain_regions')
label = '({}) {}'.format(probe['insertion_number'], label)
_, period_starts = _get_trial_event_times(
['sample', 'delay', 'go'], units, 'good_noearlylick_hit')
# ---- compute CD projected PSTH ----
_, proj_contra_trial, proj_ipsi_trial, time_stamps, hemi = psth.compute_CD_projected_psth(
units.fetch('KEY'), time_period=time_period)
# ---- save projection results ----
probe_proj[pid] = (proj_contra_trial, proj_ipsi_trial,
time_stamps, label, hemi)
# ---- generate fig with CD plot for this probe ----
fig, ax = plt.subplots(1, 1, figsize=(6, 6))
_plot_with_sem(proj_contra_trial, time_stamps, ax=ax, c='b')
_plot_with_sem(proj_ipsi_trial, time_stamps, ax=ax, c='r')
# cosmetic
for x in period_starts:
ax.axvline(x=x, linestyle='--', color='k')
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)
ax.set_ylabel('CD projection (a.u.)')
ax.set_xlabel('Time (s)')
ax.set_title(label)
fig.tight_layout()
# ---- plot this fig into the main figure ----
buf = io.BytesIO()
fig.savefig(buf, format='png')
buf.seek(0)
axs[pid, pid].imshow(Image.open(buf))
buf.close()
plt.close(fig)
# ---- Plot probe-pair correlation ----
for p1, p2 in itertools.combinations(probe_proj.keys(), r=2):
proj_contra_trial_g1, proj_ipsi_trial_g1, time_stamps, label_g1, p1_hemi = probe_proj[
p1]
proj_contra_trial_g2, proj_ipsi_trial_g2, time_stamps, label_g2, p2_hemi = probe_proj[
p2]
labels = [label_g1, label_g2]
# plot trial CD-endpoint correlation
p_start, p_end = time_period
contra_cdend_1 = proj_contra_trial_g1[:,
np.logical_and(
time_stamps >=
p_start, time_stamps
< p_end)].mean(
axis=1)
ipsi_cdend_1 = proj_ipsi_trial_g1[:,
np.logical_and(
time_stamps >= p_start,
time_stamps < p_end
)].mean(axis=1)
if p1_hemi == p2_hemi:
contra_cdend_2 = proj_contra_trial_g2[:,
np.logical_and(
time_stamps >=
p_start,
time_stamps <
p_end)].mean(
axis=1)
ipsi_cdend_2 = proj_ipsi_trial_g2[:,
np.logical_and(
time_stamps >=
p_start, time_stamps
< p_end)].mean(
axis=1)
else:
contra_cdend_2 = proj_ipsi_trial_g2[:,
np.logical_and(
time_stamps >=
p_start,
time_stamps < p_end
)].mean(axis=1)
ipsi_cdend_2 = proj_contra_trial_g2[:,
np.logical_and(
time_stamps >=
p_start,
time_stamps < p_end
)].mean(axis=1)
c_df = pd.DataFrame([contra_cdend_1, contra_cdend_2]).T
c_df.columns = labels
c_df['trial-type'] = 'contra'
i_df = pd.DataFrame([ipsi_cdend_1, ipsi_cdend_2]).T
i_df.columns = labels
i_df['trial-type'] = 'ipsi'
df = c_df.append(i_df)
# remove NaN trial - could be due to some trials having no spikes
non_nan = ~np.logical_or(
np.isnan(df[labels[0]]).values,
np.isnan(df[labels[1]]).values)
df = df[non_nan]
fig = plt.figure(figsize=(6, 6))
jplot = _jointplot_w_hue(data=df,
x=labels[0],
y=labels[1],
hue='trial-type',
colormap=['b', 'r'],
figsize=(8, 6),
fig=fig,
scatter_kws=None)
# ---- plot this fig into the main figure ----
buf = io.BytesIO()
fig.savefig(buf, format='png')
buf.seek(0)
axs[p1, p2].imshow(Image.open(buf))
buf.close()
plt.close(fig)
else:
# ---- Plot Single-Probe Coding Direction ----
probe = probe_keys[0]
units = ephys.Unit & probe
label = (ephys.ProbeInsertion & probe).aggr(
ephys.ProbeInsertion.RecordableBrainRegion.proj(
brain_region='CONCAT(hemisphere, " ", brain_area)'),
brain_regions='GROUP_CONCAT(brain_region SEPARATOR", ")'
).fetch1('brain_regions')
unit_characteristic_plot.plot_coding_direction(
units, time_period=time_period, label=label, axs=axs)
# ---- Save fig and insert ----
fn_prefix = f'{water_res_num}_{sess_date}_'
fig_dict = save_figs((fig1, ), ('coding_direction', ), sess_dir,
fn_prefix)
plt.close('all')
self.insert1({**key, **fig_dict, 'cd_probe_count': len(probe_keys)})
def plot_paired_coding_direction(unit_g1, unit_g2, labels=None, time_period=None):
"""
Plot trial-to-trial CD-endpoint correlation between CD-projected trial-psth from two unit-groups (e.g. two brain regions)
Note: coding direction is calculated on selective units, contra vs. ipsi, within the specified time_period
"""
_, proj_contra_trial_g1, proj_ipsi_trial_g1, time_stamps, unit_g1_hemi = psth.compute_CD_projected_psth(
unit_g1.fetch('KEY'), time_period=time_period)
_, proj_contra_trial_g2, proj_ipsi_trial_g2, time_stamps, unit_g2_hemi = psth.compute_CD_projected_psth(
unit_g2.fetch('KEY'), time_period=time_period)
# get event start times: sample, delay, response
period_names, period_starts = _get_trial_event_times(['sample', 'delay', 'go'], unit_g1, 'good_noearlylick_hit')
if labels:
assert len(labels) == 2
else:
labels = ('unit group 1', 'unit group 2')
# plot projected trial-psth
fig, axs = plt.subplots(1, 2, figsize=(16, 6))
_plot_with_sem(proj_contra_trial_g1, time_stamps, ax=axs[0], c='b')
_plot_with_sem(proj_ipsi_trial_g1, time_stamps, ax=axs[0], c='r')
_plot_with_sem(proj_contra_trial_g2, time_stamps, ax=axs[1], c='b')
_plot_with_sem(proj_ipsi_trial_g2, time_stamps, ax=axs[1], c='r')
# cosmetic
for ax, label in zip(axs, labels):
for x in period_starts:
ax.axvline(x=x, linestyle = '--', color = 'k')
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)
ax.set_ylabel('CD projection (a.u.)')
ax.set_xlabel('Time (s)')
ax.set_title(label)
# plot trial CD-endpoint correlation - if 2 unit-groups are from 2 hemispheres,
# then contra-ipsi definition is based on the first group
p_start, p_end = time_period
contra_cdend_1 = proj_contra_trial_g1[:, np.logical_and(time_stamps >= p_start, time_stamps < p_end)].mean(axis=1)
ipsi_cdend_1 = proj_ipsi_trial_g1[:, np.logical_and(time_stamps >= p_start, time_stamps < p_end)].mean(axis=1)
if unit_g1_hemi == unit_g1_hemi:
contra_cdend_2 = proj_contra_trial_g2[:, np.logical_and(time_stamps >= p_start, time_stamps < p_end)].mean(axis=1)
ipsi_cdend_2 = proj_ipsi_trial_g2[:, np.logical_and(time_stamps >= p_start, time_stamps < p_end)].mean(axis=1)
else:
contra_cdend_2 = proj_ipsi_trial_g2[:, np.logical_and(time_stamps >= p_start, time_stamps < p_end)].mean(axis=1)
ipsi_cdend_2 = proj_contra_trial_g2[:, np.logical_and(time_stamps >= p_start, time_stamps < p_end)].mean(axis=1)
c_df = pd.DataFrame([contra_cdend_1, contra_cdend_2]).T
c_df.columns = labels
c_df['trial-type'] = 'contra'
i_df = pd.DataFrame([ipsi_cdend_1, ipsi_cdend_2]).T
i_df.columns = labels
i_df['trial-type'] = 'ipsi'
df = c_df.append(i_df)
jplot = _jointplot_w_hue(data=df, x=labels[0], y=labels[1], hue= 'trial-type', colormap=['b', 'r'],
figsize=(8, 6), fig=None, scatter_kws=None)
jplot['fig'].show()
return fig