)
ix = ~e6194_obj.meta.cell_line.isin([
'NA07057',
'HCT116',
'HEL46.11',
])
e6194_obj.filter_samples(ix)
e6194_obj.meta.insert(1, 'array_type', 'EPIC')
# GSE110544 (Banovich et al.; iPSC lines) (EPIC)
# banov_obj = loader.load_reference('gse110544', norm_method=norm_method)
# banov_obj.meta.insert(1, 'array_type', 'EPIC')
# HipSci data
hip_epic_obj = loader.hipsci(norm_method=norm_method,
array_type='epic',
n_sample=12)
# combine all data
obj = loader.loader.MultipleBatchLoader(
# [me_obj, encode_epic_obj, e6194_obj, hip_epic_obj, banov_obj]
[me_obj, encode_epic_obj, e6194_obj, hip_epic_obj])
meta = obj.meta
dat_m = process.m_from_beta(obj.data)
this_anno = anno.loc[dat_m.index]
dmr_clusters = compute_dmr_clusters(this_anno, dmr_params)
ipsc_ref_names_6194 = ['HEL139', 'HEL140', 'HEL141']
ipsc_ref_names_6194_n1 = [
'HEL139.2_p17', 'HEL139.5_p14', 'HEL139.8_p13', 'HEL140.1_p12',
norm_method=norm_method,
samples=our_samples)
nazor_ldr = loader.load_reference('GSE31848', norm_method=norm_method)
ix = nazor_ldr.meta.index.str.contains(r'(ES__WA)|(iPS__HDF)')
ix = ix & (~nazor_ldr.meta.index.str.contains(r'HDF51IPS7')
) # this sample is an outlier, so remove it now
nazor_ldr.filter_samples(ix)
# Zhou et al.: lots of samples here, but we'll only keep 2 x ESC lines
zhou_ldr = loader.load_reference('GSE92462_450K', norm_method=norm_method)
ix = zhou_ldr.meta.index.str.contains(r'^H[19]ES')
zhou_ldr.filter_samples(ix)
hip_epic_ldr = loader.hipsci(norm_method=norm_method,
n_sample=n_hipsci,
array_type='epic')
## FIXME: this is required to avoid a BUG where the meta column gets renamed to batch_1 in all other loaders
hip_epic_ldr.meta.drop('batch', axis=1, inplace=True)
# Weltner et al. (E-MTAB-6194)
e6194_ldr = loader.load_reference('E-MTAB-6194', norm_method=norm_method)
ix = ~e6194_ldr.meta.cell_line.isin(
['NA07057', 'HCT116', 'HEL46.11', 'CCD-1112Sk (CRL-2429)'])
e6194_ldr.filter_samples(ix)
refs = [('Kim et al.',
loader.gse38216(
norm_method=norm_method,
samples=['H9 ESC 1', 'H9 ESC 2', 'H9 NPC 1', 'H9 NPC 2'])),
('Morey et al.',
# r.batch_id = bid
# r.meta.index = ["%s_%s" % (t, bid) for t in r.meta.index]
# r.data.columns = r.meta.index
# ref_obj = loader.loader.MultipleBatchLoader([t[1] for t in refs])
# ref_meta = ref_obj.meta.loc[ref_obj.meta.index.str.contains('ESC')]
# ref_data = ref_data.loc[:, ref_meta.index]
# combine all loaders
# this will reduce the probe list to the intersection (i.e. 450K)
ref_obj = loader.loader.MultipleBatchLoader(refs)
ref_meta = ref_obj.meta
ref_data = ref_obj.data.dropna()
# HipSci data
hip_epic_ldr = loader.hipsci(norm_method=norm_method, n_sample=12, array_type='epic')
hip_epic_meta = hip_epic_ldr.meta
hip_epic_data = hip_epic_ldr.data
# hip_450k_meta, hip_450k_data = loader.hipsci(norm_method=norm_method, n_sample=30, array_type='450k')
hip_ldr = loader.hipsci(norm_method=norm_method, n_sample=12, array_type='all')
hip_meta = hip_ldr.meta
hip_data = hip_ldr.data
hip_meta.batch = ["HipSci (%s)" % t for t in hip_meta.array_type]
# clustering genome-wide
# iPSC, FB, ESC
# mix of HipSci samples by array_type
meta, dat = combine_data_meta(