def pairwise_comp(data, cty_prop, prop_list, params, sig_level=0.05):
"""
Pairwise comparison of parameters between cell-types
"""
diff_param_list = []
p_val_list = []
for param in params:
for comb in combinations(prop_list, 2):
cty_x, cty_y = comb
paramx = data.loc[data[cty_prop] == cty_x, param].values
paramy = data.loc[data[cty_prop] == cty_y, param].values
_, p_val_x = mannwhitneyu(paramx, paramy, alternative='less')
_, p_val_y = mannwhitneyu(paramy, paramx, alternative='less')
comp_type = '%s<%s' % (
cty_x, cty_y) if p_val_x < p_val_y else '%s<%s' % (cty_y, cty_x)
p_val = min(p_val_x, p_val_y)
sig_dict = {'Comp_type': comp_type,
'param': param}
diff_param_list.append(sig_dict)
p_val_list.append(p_val)
# FDR correction for multiple comparison
_, p_val_corrected = fdrcorrection(p_val_list)
diff_param_df = pd.DataFrame(diff_param_list)
diff_param_df['p_val'] = p_val_corrected
diff_param_df['sig_level'] = diff_param_df['p_val'].apply(
lambda x: man_utils.pval_to_sig(x))
return diff_param_df
def draw_significance(mu_dist1, mu_dist2, pval, ax, height_offset=.1):
sig_text = man_utils.pval_to_sig(pval)
bin_dist1, _ = np.histogram(mu_dist1, density=True, bins=10)
bin_dist2, _ = np.histogram(mu_dist2, density=True, bins=10)
mean_dist1 = np.mean(mu_dist1)
mean_dist2 = np.mean(mu_dist2)
y_height1 = np.max(bin_dist1)
y_height2 = np.max(bin_dist2)
max_y_height = (1 + height_offset) * np.max([y_height1, y_height2])
ax.plot([mean_dist1, mean_dist1, mean_dist2, mean_dist2],
[y_height1, max_y_height, max_y_height, y_height2],
color='k')
ax.text((mean_dist1 + mean_dist2) * .5,
max_y_height,
sig_text,
ha='center',
va='bottom',
color='k')
return ax
def sig_test(feature_df):
diff_ephys_df = []
p_val_list = []
for efeat in all_renamed_feats:
for comb in combinations(inh_subclasses, 2): # 2 for pairs, 3 for triplets, etc
subclass_x_idx, subclass_y_idx = comb
subclass_x_idx_efeat = feature_df.loc[feature_df.ttype == subclass_x_idx, efeat].values
subclass_y_idx_efeat = feature_df.loc[feature_df.ttype == subclass_y_idx, efeat].values
_, p_val = mannwhitneyu(subclass_x_idx_efeat,
subclass_y_idx_efeat, alternative='two-sided')
comp_type = '%s~%s' % (subclass_x_idx, subclass_y_idx)
sig_dict = {'comp_type': comp_type,
'feature': efeat}
diff_ephys_df.append(sig_dict)
p_val_list.append(p_val)
_, p_val_corrected = fdrcorrection(p_val_list)
diff_ephys_df = pd.DataFrame(diff_ephys_df)
diff_ephys_df['p_val'] = p_val_corrected
diff_ephys_df['sig_level'] = diff_ephys_df['p_val'].apply(lambda x: man_utils.pval_to_sig(x))
return diff_ephys_df
inter_mask_blkdiag = np.kron(
inter_mask_cell, np.ones((len(unique_cell_ids), len(unique_cell_ids))))
masked_intra_conductance = np.multiply(conductance_dist_matrix,
intra_mask_blkdiag).flatten()
masked_intra_conductance = masked_intra_conductance[
masked_intra_conductance != 0]
masked_inter_conductance = np.multiply(conductance_dist_matrix,
inter_mask_blkdiag).flatten()
masked_inter_conductance = masked_inter_conductance[
masked_inter_conductance != 0]
_, p_less_ = mannwhitneyu(masked_intra_conductance,
masked_inter_conductance,
alternative='less')
sig_text = man_utils.pval_to_sig(p_less_)
intra_dist[subclass_] = masked_intra_conductance.tolist()
inter_dist[subclass_] = masked_inter_conductance.tolist()
axis_fontsize = 14
tick_fontsize = 12
legend_fontsize = axis_fontsize
ax[ii // 2, ii % 2] = sns.distplot(intra_dist[subclass_],
norm_hist=True,
ax=ax[ii // 2, ii % 2],
hist_kws={'label': 'intra'},
color=intra_cell_intra_class_col)
ax[ii // 2, ii % 2] = sns.distplot(inter_dist[subclass_],
norm_hist=True,
ax=ax[ii // 2, ii % 2],
cre_x, cre_y) if p_val_x < p_val_y else '%s<%s' % (cre_y, cre_x)
p_val = min(p_val_x, p_val_y)
sig_dict = {
'Comp_type': comp_type,
'gene': gene_,
}
diff_gene_expression_df.append(sig_dict)
p_val_list.append(p_val)
# FDR correction @5%
_, p_val_corrected = fdrcorrection(p_val_list)
diff_gene_expression_df = pd.DataFrame(diff_gene_expression_df)
diff_gene_expression_df['p_val'] = p_val_corrected
diff_gene_expression_df['sig_level'] = diff_gene_expression_df['p_val'].apply(
lambda x: man_utils.pval_to_sig(x))
diff_gene_expression_df = diff_gene_expression_df.loc[
diff_gene_expression_df.sig_level != 'n.s.', ]
gene_sig_grouped = diff_gene_expression_df.groupby('gene')
exc_expression_melted = pd.melt(exc_expression_data,
id_vars=['sample_id', 'Cre_line'],
value_vars=h_genes,
var_name='gene',
value_name='cpm')
exc_expression_melted['Cre_gene'] = exc_expression_melted.apply(
lambda x: x.gene + '.' + x.Cre_line, axis=1)
comp_types = exc_expression_melted['Cre_gene'].unique().tolist()
data_types = exc_expression_melted.gene.unique().tolist()
for comb in combinations(inh_lines, 2): # 2 for pairs, 3 for triplets, etc
cre_x,cre_y = comb
cre_x_efeat = select_spiking_df.loc[select_spiking_df.Cre_line == cre_x,efeat].values
cre_y_efeat = select_spiking_df.loc[select_spiking_df.Cre_line == cre_y,efeat].values
_,p_val = mannwhitneyu(cre_x_efeat,cre_y_efeat,alternative='two-sided')
comp_type = '%s~%s'%(cre_x,cre_y)
sig_dict = {'comp_type' : comp_type,
'feature': efeat}
diff_ephys_df.append(sig_dict)
p_val_list.append(p_val)
_,p_val_corrected = fdrcorrection(p_val_list)
diff_ephys_df = pd.DataFrame(diff_ephys_df)
diff_ephys_df['p_val'] = p_val_corrected
diff_ephys_df['sig_level'] = diff_ephys_df['p_val'].apply(lambda x: man_utils.pval_to_sig(x))
spiking_melt_df = pd.melt(select_spiking_df,id_vars=['Cell_id','Cre_line'],
value_vars=all_renamed_feats,var_name='features',value_name='value')
# filtered ME cells
filtered_me_inh_cells = utility.load_pickle(filtered_me_inh_cells_filename)
spiking_melt_df = spiking_melt_df.loc[spiking_melt_df.Cell_id.isin(filtered_me_inh_cells),]
ylim_list = [1.6, 60,220, 2.0]
sns.set(style='whitegrid')
fig,ax = plt.subplots(1,len(all_renamed_feats),sharey=False,figsize=(12,3))
for ii,feat_ in enumerate(all_renamed_feats):
data = spiking_melt_df.loc[spiking_melt_df.features == feat_,]
'Model': sag_features_model,
# 'Rbp4like':sag_perturbed_Rbp4,
# 'Nr5like':sag_perturbed_Nr5
}
sig_dict_list = []
for type_, data_ in data_dict.items():
feat_Rbp4 = data_.loc[data_.Cre_line == "Rbp4-Cre_KL100",
select_sag_feature].values
feat_Nr5 = data_.loc[data_.Cre_line == "Nr5a1-Cre",
select_sag_feature].values
_, p_val = mannwhitneyu(feat_Nr5, feat_Rbp4, alternative='two-sided')
sig_dict_list.append({
'data_type': type_,
'sig_level': man_utils.pval_to_sig(p_val),
'Comp_type': "Nr5a1-Cre~Rbp4-Cre_KL100"
})
sig_df = pd.DataFrame(sig_dict_list)
ephys_sig_group = sig_df.groupby('data_type')
sig_vars = sig_df.data_type.tolist()
sag_features_all = pd.concat(
[
sag_features_exp,
sag_features_exp_selected,
sag_features_model,
#sag_perturbed_Rbp4,sag_perturbed_Nr5
],
sort=False)