def test_read(self):
in_gmt = gmt.read(os.path.join(FUNCTIONAL_TESTS_DIR, "test.gmt"))
self.assertEqual(len(self.example_gmt), len(in_gmt))
self.assertEqual(self.example_gmt[0], in_gmt[0])
self.assertEqual(self.example_gmt[1], in_gmt[1])
with self.assertRaises(AssertionError) as e:
gmt.read(os.path.join(FUNCTIONAL_TESTS_DIR, "test_bad.gmt"))
self.assertIn("3 tab-delimited items. line_num: 0", str(e.exception))
with self.assertRaises(AssertionError) as e:
gmt.read(os.path.join(FUNCTIONAL_TESTS_DIR, "test_bad2.gmt"))
self.assertIn("same set. line_num: 1", str(e.exception))
def test_write(self): out_path = os.path.join(FUNCTIONAL_TESTS_DIR, "test_write.gmt") gmt.write(self.example_gmt, out_path) self.assertTrue(os.path.exists(out_path)) read_back_in = gmt.read(out_path) self.assertEqual(len(self.example_gmt), len(read_back_in)) self.assertEqual(self.example_gmt[0], read_back_in[0]) self.assertEqual(self.example_gmt[1], read_back_in[1]) # Cleanup os.remove(out_path)
def wtks(gct,
metadata,
outfolder,
gmt_path='/Users/elemire/Workspace/merino/full_sensitivities.gmt',
group_col='weave_prefix'):
gmt_file = gmt.read(gmt_path)
sensitivity_map = reformat_gmt(gmt_file)
metadata = metadata.loc[gct.data_df.columns]
expected_sensitivity_ranks = []
for rep in metadata[group_col].dropna().unique():
if not os.path.exists(os.path.join(outfolder, rep)):
os.mkdir(os.path.join(outfolder, rep))
os.mkdir(os.path.join(outfolder, rep, 'enrichment_ecdf'))
os.mkdir(os.path.join(outfolder, rep, 'mountain_plots'))
rep_folder = os.path.join(outfolder, rep)
ids = metadata[metadata[group_col] == rep].index
data = gct.data_df[ids]
data.index = [str(x) for x in data.index]
col_meta = metadata.loc[ids]
marks = {}
s_qi_map = make_sqi_map(sensitivity_map, data=data, col_meta=col_meta)
for key in s_qi_map:
#print key.split('_')[0]
if key.split('_')[0] in col_meta['pert_id'].tolist():
if len(s_qi_map[key][0]) > 0 and len(s_qi_map[key][1]) > 0:
#calculate enrichment score
sensitivity_score, cumsum = compute_wtcs.compute_wtcs(
pd.Series(s_qi_map[key][0]),
pd.Series(s_qi_map[key][1]))
#make mountain plot of enrichment score
plot_enrichment_score.plot_enrichment_score(
sensitivity_score,
cumsum,
title='Enrichment Score of {}, Set Size = {}'.format(
col_meta[col_meta['pert_id'] == key.split('_')[0]]
['pert_iname'][0], len(s_qi_map[key][1])),
outfile=os.path.join(
rep_folder, 'mountain_plots', '{}_{}_dose'.format(
col_meta[col_meta['pert_id'] == key.split('_')
[0]]['pert_iname'][0],
key.split('_')[1]) + '.png'))
bortezomib_scores = []
dmso_scores = []
pos_dex = col_meta[col_meta['pert_type'] ==
'trt_poscon'].index.tolist()
neg_dex = col_meta[col_meta['pert_type'] ==
'ctl_vehicle'].index.tolist()
bortez = data[pos_dex]
dmso = data[neg_dex]
brd = key.split('_')[0]
rids = sensitivity_map[brd]
################################
for column in bortez:
temp_s = bortez[column].dropna()
temp_qi = [
thing for thing in np.where(
temp_s.dropna().index.isin(rids))[0].tolist()
]
if (len(temp_s) - len(temp_qi)) < 1:
continue
y, x = compute_wtcs.compute_wtcs(
temp_s, pd.Series(temp_qi))
bortezomib_scores.append(y)
for column in dmso:
temp_s = dmso[~dmso.index.isin(invariants
)][column].dropna()
temp_qi = [
thing for thing in np.where(
temp_s.dropna().index.isin(rids))[0].tolist()
]
if (len(temp_s) - len(temp_qi)) < 1:
continue
y, x = compute_wtcs.compute_wtcs(
temp_s, pd.Series(temp_qi))
dmso_scores.append(y)
all_scores = []
for column in data:
temp_s = data[~data.index.isin(invariants
)][column].dropna()
temp_qi = [
thing for thing in np.where(
temp_s.dropna().index.isin(rids))[0].tolist()
]
if (len(temp_s) - len(temp_qi)) < 1:
continue
y, x = compute_wtcs.compute_wtcs(
temp_s, pd.Series(temp_qi))
all_scores.append(y)
ecdf = ECDF(all_scores)
mark = bisect.bisect_left(ecdf.x, sensitivity_score)
if mark == len(ecdf.y):
mark -= 1
marks[col_meta[col_meta['pert_id'] == key.split('_')
[0]]['pert_iname'][0] + '_' +
str(key.split('_')[1])] = mark
poscons_x = []
poscons_y = []
for pos in bortezomib_scores:
y = bisect.bisect_left(ecdf.x, pos)
poscons_x.append(pos)
poscons_y.append(ecdf.y[y - 1])
neg_x = []
neg_y = []
for neg in dmso_scores:
y = bisect.bisect_left(ecdf.x, neg)
neg_x.append(neg)
neg_y.append(ecdf.y[y - 1])
plt.figure()
plt.plot(ecdf.x, ecdf.y)
plt.scatter(poscons_x,
poscons_y,
marker='o',
color='g',
label='Bortezomib')
# plt.scatter(neg_x, neg_y, marker='o', color='b', label='DMSO')
plt.scatter(sensitivity_score,
ecdf.y[mark],
marker='o',
color='r',
label='Sensitivity {}_dose'.format(
key.split('_')[1]))
plt.xlabel(
'Enrichment Score for {} Sensitivities Compound Rank = {}'
.format(
col_meta[col_meta['pert_id'] == key.split('_')[0]]
['pert_iname'][0], mark))
plt.ylabel('Fraction of Compounds')
plt.title(
'ECDF of Enrichment Score by Compound, Set Size = {}'.
format(len(s_qi_map[key][1])))
axes = plt.gca()
axes.set_xlim(
[np.nanmin(ecdf.x[ecdf.x != -np.inf]),
max(ecdf.x)])
axes.set_ylim([0, 1])
axes.legend(bbox_to_anchor=(0., 0.8, 0.8, .102),
loc=3,
borderaxespad=0.)
plt.savefig(
os.path.join(
rep_folder, 'enrichment_ecdf',
'{}_Competitive_Enrichment_ECDF_{}_dose.png'.
format(
col_meta[col_meta['pert_id'] == key.split('_')
[0]]['pert_iname'][0],
key.split('_')[1])))
plt.clf()
plate_summary = pd.Series(marks)
# plate_summary.set_index('det_plate', inplace=True)
plate_summary.rename(rep, inplace=True)
pd.DataFrame(plate_summary).to_csv(os.path.join(
rep_folder, '{}_expected_sensitivity_ranks.txt'.format(rep)),
sep='\t')
expected_sensitivity_ranks.append(marks)
marks['det_plate'] = rep
summary = pd.DataFrame(expected_sensitivity_ranks)
summary.set_index('det_plate', inplace=True)
summary.to_csv(os.path.join(outfolder, 'expected_sensitivity_ranks.txt'),
sep='\t')