def parse_data(res, excitatory=True):
list_odor_on = res['DAQ_O_ON_F']
list_odor_off = res['DAQ_O_OFF_F']
list_water_on = res['DAQ_W_ON_F']
list_data = res['data']
for i, data in enumerate(list_data):
start_time = time.time()
p_list = []
dff_list = []
for cell_data in data:
baseline = cell_data[:, config.baseline_start:list_odor_on[i] -
config.baseline_end]
baseline = baseline.flatten()
data_trial_time_window = _rolling_window(
cell_data, window=condition_config.p_window)
data_time_trial_window = np.transpose(data_trial_time_window,
[1, 0, 2])
data_time_pixels = data_time_trial_window.reshape(
data_time_trial_window.shape[0], -1)
#statistical significance
if excitatory:
f = lambda y: mannwhitneyu(
baseline, y, use_continuity=True, alternative='less')[-1]
else:
f = lambda y: mannwhitneyu(
baseline, y, use_continuity=True, alternative='greater')[-1
]
p = []
for y in data_time_pixels:
try:
out = f(y)
except:
out = 1
p.append(out)
while (len(p) < cell_data.shape[-1]):
p.append(p[-1])
#magnitude
dff = psth.psth_helper.subtract_baseline(
cell_data, config.baseline_start,
list_odor_on[i] - config.baseline_end)
dff = np.mean(dff, axis=0)
p_list.append(np.array(p))
dff_list.append(np.array(dff))
res['dff'].append(np.array(dff_list))
res['p'].append(np.array(p_list))
elapsed = time.time() - start_time
print('analyzed mouse {}, day {}, odor {}, in {} seconds'.format(
res['mouse'][i], res['day'][i], res['odor'][i], elapsed))
for key, val in res.items():
res[key] = np.array(val)
# res.pop('data')
name_str = 'dict'
if not excitatory:
name_str += '_I'
fio.save_pickle(save_path=save_path, save_name=name_str, data=res)
def do_PCA(condition, PCAConfig, data_path, save_path):
'''
Run PCA experiment. For each mouse, data should be in format of (trials * time) X neurons,
stacked for all odor conditions
:param condition:
:param PCAConfig:
:param data_path:
:param save_path:
:return:
'''
data_pathnames = sorted(
glob.glob(os.path.join(data_path, '*' + Config.mat_ext)))
config_pathnames = sorted(
glob.glob(os.path.join(data_path, '*' + Config.cons_ext)))
list_of_all_data = np.array([Config.load_mat_f(d) for d in data_pathnames])
list_of_all_cons = np.array(
[Config.load_cons_f(d) for d in config_pathnames])
mouse_names_per_file = np.array(
[cons.NAME_MOUSE for cons in list_of_all_cons])
mouse_names, list_of_mouse_ix = np.unique(mouse_names_per_file,
return_inverse=True)
if mouse_names.size != len(condition.paths):
raise ValueError(
"res has {0:d} mice, but filter has {1:d} mice".format(
mouse_names.size, len(condition.paths)))
for i, mouse_name in enumerate(mouse_names):
ix = mouse_name == mouse_names_per_file
start_day = 0
list_of_cons = list_of_all_cons[ix][start_day:]
list_of_data = list_of_all_data[ix][start_day:]
for cons in list_of_cons:
assert cons.NAME_MOUSE == mouse_name, 'Wrong mouse file!'
if hasattr(condition, 'odors'):
odor = condition.odors[i]
csp = condition.csp[i]
else:
odor = condition.dt_odors[i] + condition.pt_csp[i]
csp = condition.dt_csp[i] + condition.pt_csp[i]
name = list_of_cons[0].NAME_MOUSE
out, variance_explained = PCA(list_of_cons, list_of_data, odor, csp,
PCAConfig)
fio.save_pickle(save_path=save_path, save_name=name, data=out)
print("Analyzed: {}. Variance explained: {}".format(
name, variance_explained))
def organizer_test_split(condition, decodeConfig, data_path, save_path):
'''
:param condition:
:param decodeConfig:
:param data_path:
:param save_path:
:return:
'''
data_pathnames = sorted(
glob.glob(os.path.join(data_path, '*' + Config.mat_ext)))
config_pathnames = sorted(
glob.glob(os.path.join(data_path, '*' + Config.cons_ext)))
list_of_all_data = np.array([Config.load_mat_f(d) for d in data_pathnames])
list_of_all_cons = np.array(
[Config.load_cons_f(d) for d in config_pathnames])
mouse_names_per_file = np.array(
[cons.NAME_MOUSE for cons in list_of_all_cons])
mouse_names, list_of_mouse_ix = np.unique(mouse_names_per_file,
return_inverse=True)
if mouse_names.size != len(condition.paths):
raise ValueError(
"res has {0:d} mice, but filter has {1:d} mice".format(
mouse_names.size, len(condition.paths)))
start_days = decodeConfig.start_day
end_days = decodeConfig.end_day
days_per_mouse = []
for x, y in zip(start_days, end_days):
days_per_mouse.append(np.arange(x, y + 1))
print(days_per_mouse)
for i, mouse_name in enumerate(mouse_names):
start_time = time.time()
ix = mouse_name == mouse_names_per_file
list_of_cons_ = list_of_all_cons[ix]
list_of_data_ = list_of_all_data[ix]
for cons in list_of_cons_:
assert cons.NAME_MOUSE == mouse_name, 'Wrong mouse file!'
if len(days_per_mouse[i]):
list_of_cons = list_of_cons_[days_per_mouse[i]]
list_of_data = list_of_data_[days_per_mouse[i]]
cons = list_of_cons[0]
cons_dict = cons.__dict__
for key, value in cons_dict.items():
if isinstance(value, list) or isinstance(value, np.ndarray):
pass
else:
setattr(decodeConfig, key, value)
if hasattr(condition, 'odors'):
odor = condition.odors[i]
csp = condition.csp[i]
else:
odor = condition.dt_odors[i]
csp = condition.dt_csp[i]
cons_odors = [cons.ODOR_UNIQUE for cons in list_of_cons]
ix = [
i for i, unique_odors in enumerate(cons_odors)
if odor[0] in unique_odors
]
list_of_cons = [list_of_cons[i] for i in ix]
list_of_data = [list_of_data[i] for i in ix]
scores_res = decoding.test_split(list_of_cons, list_of_data, odor,
csp, decodeConfig)
name = cons.NAME_MOUSE
fio.save_json(save_path=save_path,
save_name=name,
config=decodeConfig)
fio.save_pickle(save_path=save_path,
save_name=name,
data=scores_res)
print("Analyzed: {0:s} in {1:.2f} seconds".format(
name,
time.time() - start_time))
imdir = os.path.join(d, 'python_reg.tif')
fi.imsave(imdir, mean_ims.astype('uint16'), photometric='minisblack')
for i, im in enumerate(mean_ims):
plt.imshow(im, cmap='gray')
plt.xticks([])
plt.yticks([])
plot._easy_save(d, str(i))
#saving
p, experiment = os.path.split(d)
p, mouse = os.path.split(p)
save_dir = os.path.join(data_directory, condition.name)
res = defaultdict(list)
res['within_day_corrs'].append(within_day_corrs)
res['within_day_corrs_average'].append(within_day_corr_average)
res['within_day_crisp'].append(within_day_crisp)
res['within_day_crisp_average'].append(np.mean(within_day_crisp))
res['across_day_mean_corrs'].append(across_day_mean_corrs)
res['across_day_mean_corrs_average'].append(across_day_mean_corrs_average)
# res['within_day_mean_corrs'].append(within_day_mean_corrs)
res['mouse'].append(mouse)
res['experiment'].append(condition.name)
for k, v in res.items():
res[k] = np.array(v)
if k in ['mouse', 'within_day_crisp', 'across_day_mean_corrs_average']:
print('{}: {}'.format(k, v))
file_io.save_pickle(save_path=save_dir, save_name=mouse, data=res)