def KS(adata):
adata_1 = adata[adata.obs.loc[:, 'DENV_reads'] == 0, ]
adata_2 = adata[adata.obs.loc[:, 'DENV_reads'] != 0, ]
results = anndataks.compare(adata_1, adata_2, log1p=2)
results_sort = results.sort_values(by='statistic', ascending=False)
# results_sort_2 = results_sort[(results_sort.pvalue < 0.05)]
return (results_sort)
def test_compare_sparse():
import scipy.sparse
X1 = np.array([
[0, 1],
[1, 2],
[2, 3],
[3, 4],
[3, 5],
])
X2 = np.array([
[0, 1],
[6, 2],
[6, 3],
[6, 5],
])
# Make sparse
X1 = scipy.sparse.csc_matrix(X1)
X2 = scipy.sparse.csc_matrix(X2)
adata1 = anndata.AnnData(X=X1)
adata2 = anndata.AnnData(X=X2)
adata1.var_names = ['Gene1', 'Gene2']
adata2.var_names = ['Gene1', 'Gene2']
anndataks.rc['use_experimental_ks_2samp'] = True
ress = anndataks.compare(adata1, adata2, log1p=False)
ress_exp = pd.DataFrame(
[[0.75, 1.5, 0.142857, 1.485427, 2.459432, 0.974005],
[-0.15, 3, 1.000000, 2.000000, 1.906891, -0.093109]],
columns=['statistic', 'value', 'pvalue', 'avg1', 'avg2', 'log2_fold_change'],
index=adata1.var_names,
)
assert((ress.shape == ress_exp.shape))
assert((np.abs(ress.values - ress_exp.values) < 1e-3).all())
anndataks.rc['use_experimental_ks_2samp'] = False
ress = anndataks.compare(adata1, adata2, log1p=False)
ress_exp = pd.DataFrame(
[[0.75, 0.142857, 1.485427, 2.459432, 0.974005],
[0.15, 1.000000, 2.000000, 1.906891, -0.093109]],
columns=['statistic', 'pvalue', 'avg1', 'avg2', 'log2_fold_change'],
index=adata1.var_names,
)
assert((ress.shape == ress_exp.shape))
assert((np.abs(ress.values - ress_exp.values) < 1e-3).all())
import anndataks
cell_types = [
'B_cells', 'Monocytes', 'NK_cells', 'Plasmablasts', 'T_cells', 'cDCs',
'pDCs'
]
conditions = ['S_dengue', 'dengue']
#sc.pp.log1p(adatag)
adata_kids = adatag[adatag.obs['dataset'] == 'child']
results = {}
for cell_type in cell_types:
adata_ct = adata_kids[adata_kids.obs['cell_type'] == cell_type]
if cell_type == 'cDCs':
adata_ct = adata_ct[~adata_ct.obs['ID'].isin(['1_140_01', '5_193_01'])]
adata_SD = adata_ct[adata_ct.obs['Condition'] == 'S_dengue']
adata_D = adata_ct[adata_ct.obs['Condition'] == 'dengue']
results[cell_type] = anndataks.compare(
adata_D,
adata_SD) # log1p=False # log2_fold_change: adata_Sd vs adata_D
res = pd.DataFrame([])
for cell_type in cell_types:
results[cell_type]['cell_type'] = [cell_type] * results[cell_type].shape[0]
res = pd.concat([res, results[cell_type]])
res.to_csv(
'/home/yike/phd/dengue/data/excels/log2_fc/S_dengue_vs_dengue/inters/inter_genes_pvalue.tsv'
)
conditions = ['S_dengue', 'dengue']
results = {}
for subct in subcts:
if subct in adata_kid.obs['cell_type_new'].astype(
'category').cat.categories:
adata_ct = adata_kid[adata_kid.obs['cell_type_new'] == subct]
elif subct in adata_kid.obs['cell_subtype_new'].astype(
'category').cat.categories:
adata_ct = adata_kid[adata_kid.obs['cell_subtype_new'] == subct]
adata_SD = adata_ct[adata_ct.obs['Condition'] == 'S_dengue']
adata_D = adata_ct[adata_ct.obs['Condition'] == 'dengue']
# while calculating ks test pvalue, the adata is log1ped, so the argument log1p=2
results[subct] = anndataks.compare(
adata_D, adata_SD, log1p=2,
mode='asymp') # log2_fold_change: adata_Sd vs adata_D
ks_res = pd.DataFrame([])
for subct in subcts:
results[subct]['cell_subtype'] = [subct] * results[subct].shape[0]
ks_res = pd.concat([ks_res, results[subct]])
ks_res.set_index(['cell_subtype', ks_res.index], inplace=True)
ks_res.to_csv(save_tabels + 'ks_pvalue.tsv', sep='\t')
ks_res = pd.read_csv(save_tabels + 'ks_pvalue.tsv',
sep='\t',
index_col=['cell_subtype', 'Unnamed: 1'])
ct_pair = pd.concat(
[ct_pair, ks_res.loc[ct_pair.index][['statistic', 'pvalue']]], axis=1)