def test_keep_stable_genes(donor_expression, thr, per, rank, stab):
out = correct.keep_stable_genes(donor_expression,
threshold=thr,
percentile=per,
rank=rank)
assert all([isinstance(f, pd.DataFrame) for f in out])
for df1, df2 in itertools.combinations(out, 2):
assert df1.shape == df2.shape
# check that `return_stability` provides expression and stability
if thr == 0 and stab:
out, stab = correct.keep_stable_genes(donor_expression,
threshold=thr,
return_stability=stab)
assert len(stab) == len(out[0].columns)
assert np.all(out[0].columns == donor_expression[0].columns)
def test_keep_stable_genes(donor_expression):
for thr, per, rank in itertools.product(np.arange(0, 1, 0.1),
[True, False], [True, False]):
out = correct.keep_stable_genes(donor_expression,
threshold=thr,
percentile=per,
rank=rank)
assert all([isinstance(f, pd.DataFrame) for f in out])
for df1, df2 in itertools.combinations(out, 2):
assert df1.shape == df2.shape
# check that `return_stability` provides expression and stability
out, stab = correct.keep_stable_genes(donor_expression,
threshold=0,
return_stability=True)
assert len(stab) == len(out[0].columns)
assert np.all(out[0].columns == donor_expression[0].columns)
def test_keep_stable_genes(donor_expression):
for thr, per, rank in itertools.product(np.arange(0, 1, 0.1),
[True, False],
[True, False]):
out = correct.keep_stable_genes(donor_expression, threshold=thr,
percentile=per, rank=rank)
assert all([isinstance(f, pd.DataFrame) for f in out])
for df1, df2 in itertools.combinations(out, 2):
assert df1.shape == df2.shape
def select_genes(dict_genes, sphere_lh, sphere_rh, n_perm=100, df_gandal=None):
"""Gene enrichment analysis.
Parameters
----------
dict_genes: dict of pd.Dataframe
Dictionary with pd.Dataframe of genes for each donor in the AHBA
dataset.
sphere_lh: BSPolyData
Sphere for left hemisphere.
sphere_rh: BSPolyData
Sphere for right hemisphere.
n_perm: int, default=None
df_gandal: pd.DataFrame, default=None
Dataframe with logFC for ASD, SCZ and BD.
Returns
-------
selected_genes: list of str
Genes with significant spatial overlap with idiosyncrasy measures.
"""
from abagen.correct import keep_stable_genes
dict_genes = get_shared_genes(dict_genes)
genes = list(dict_genes.values())[0].columns
tval_sd, pval_sd = _select_genes_one(dict_genes,
'SD',
sphere_lh,
sphere_rh,
n_perm=n_perm)
tval_dd, pval_dd = _select_genes_one(dict_genes,
'DD',
sphere_lh,
sphere_rh,
n_perm=n_perm)
mask_selected = np.logical_and(pval_sd < 0.05, pval_dd < 0.05)
selected_genes = genes[mask_selected]
# Only keep stable genes
stable_genes = keep_stable_genes(list(dict_genes.values()),
threshold=.5,
percentile=False,
rank=True)[0]
selected_genes = np.intersect1d(stable_genes.columns, selected_genes)
if df_gandal is None:
return selected_genes
df = pd.DataFrame(np.c_[tval_sd, tval_dd],
index=genes,
columns=['SD', 'DD'])
df_gandal.rename(columns={'gene_name': 'gene_symbol'}, inplace=True)
df = df.merge(df_gandal.set_index('gene_symbol'), on='gene_symbol')
df_res = pd.DataFrame('', columns=['SD', 'DD'], index=['ASD', 'SCZ', 'BD'])
for idio in ['SD', 'DD']:
y = df[[idio]]
for k in ['ASD', 'SCZ', 'BD']:
fc = f'{k}.log2FC'
x = df[[fc, 'percentage_gene_gc_content']]
x['Constant'] = 1
rlm_results = sm.RLM(y, x).fit()
tv, pv = rlm_results.tvalues[0], rlm_results.pvalues[0]
df_res.loc[k, idio] = f'{tv:.3f} ({pv:.3f})'
return selected_genes, df_res