def get_dataset(tissue, membrane_only=True):
counts = parse_counts(tissue)
if membrane_only:
go = parse_go_plasma_membrane()
genes_membrane = go[go.isin(counts.index)]
counts = counts.loc[genes_membrane]
ds = Dataset(
samplesheet=SampleSheet(cell_types),
counts_table=CountsTable(counts),
)
return ds
def parse_counts(tissue, regenerate=False):
import glob
if 'annotation glob' in config[tissue]:
glb = config[tissue]['annotation glob']
else:
glb = tissue
if regenerate:
cglbs = ('CountTable', )
else:
cglbs = ('CountTableNormalized', 'CountTable')
for cglb in cglbs:
fn_glb = '../../data/MACAtSNE/{:}{:}.csv'.format(glb, cglb)
fns = glob.glob(fn_glb)
if len(fns):
break
if len(fns) == 0:
raise IOError('Counts file not found for tissue: {:}'.format(tissue))
elif len(fns) > 1:
raise IOError(
'Several counts files found for tissue: {:}'.format(tissue))
else:
fn = fns[0]
out = pd.read_csv(fn, sep=',', index_col=0)
if '.' in out.columns[0]:
out.columns = [
'{1}_{0}'.format(*(c.split('.')[:2])) for c in out.columns
]
out.index.name = 'GeneName'
out.columns.name = 'Cell'
out = CountsTable(out)
if 'Normalized' in fn:
out._normalized = 'counts_per_million'
else:
print('Normalize counts')
out.normalize(inplace=True)
print('Log counts')
out.log(inplace=True)
print('Write normalized counts to file')
out.to_csv(fn[:-4] + 'Normalized.csv', sep=',')
return out
def ct():
from singlet.counts_table import CountsTable
return CountsTable.from_tablename('example_table_tsv').iloc[:200]
def get_dataset(tissue,
membrane_only=True,
regenerate=False,
go_contains=None,
go_exclude=None):
# Some tissues like brain were split for sorting, we merge them here
dss = []
for tissue_facs in tissues_prediction[tissue]:
cell_types, plates = parse_annotations(tissue_facs)
counts = parse_counts(tissue_facs, regenerate=regenerate)
if membrane_only:
go = parse_go_plasma_membrane().index
genes_membrane = go[go.isin(counts.index)]
counts = counts.loc[genes_membrane]
if (go_contains is not None) and (go_exclude is not None):
raise ValueError('Use either go_contains or go_exclude')
if go_contains is not None:
go = parse_go_plasma_membrane()
genes = go.index[go['GONames'].str.contains(go_contains)]
genes = np.intersect1d(genes, counts.index)
counts = counts.loc[genes]
elif go_exclude is not None:
go = parse_go_plasma_membrane()
genes = go.index[~go['GONames'].str.contains(go_exclude)]
genes = np.intersect1d(genes, counts.index)
counts = counts.loc[genes]
dss.append({'samplesheet': cell_types, 'counts': counts})
if len(dss) == 1:
ds = Dataset(
samplesheet=SampleSheet(cell_types),
counts_table=counts,
)
return ds
else:
# Merging is kind of messy because some genes are absent from either
# subtissue (grrr); I put zeroes for now, Michelle is working on the
# better solution (we have those numbers somewhere)
genes = set()
for ds in dss:
genes |= set(ds['counts'].index.values)
genes = pd.Index(sorted(genes), name=ds['counts'].index.name)
for ds in dss:
genes_missing = genes[~genes.isin(ds['counts'].index)]
for gene in genes_missing:
# The stuff is normalized, pseudocounted, and logged
ds['counts'].loc[gene] = -1.0
ds['counts'] = ds['counts'].loc[genes]
ngenes = len(genes)
ncells = sum(ds['samplesheet'].shape[0] for ds in dss)
samplesheet_all = pd.concat([ds['samplesheet'] for ds in dss], axis=0)
counts_all = pd.DataFrame(np.zeros((ngenes, ncells), float),
index=genes,
columns=samplesheet_all.index)
for ds in dss:
counts_all.loc[:,
ds['counts'].columns.values] = ds['counts'].values
counts_all = CountsTable(counts_all)
if ds['counts']._normalized:
counts_all._normalized = ds['counts']._normalized
ds = Dataset(
samplesheet=SampleSheet(samplesheet_all),
counts_table=counts_all,
)
return ds
#!/usr/bin/env python
# vim: fdm=indent
'''
author: Fabio Zanini
date: 15/08/17
content: Test CountsTable class.
'''
# Script
if __name__ == '__main__':
# NOTE: an env variable for the config file needs to be set when
# calling this script
from singlet.counts_table import CountsTable
ct = CountsTable.from_tablename('example_table_tsv')
print('Test binning of CountsTable')
ct = ct.iloc[:200]
ct.bin(result='index', inplace=True)
assert (ct.values.max() == 4)
print('Done!')
def test_initialize():
from singlet.counts_table import CountsTable
ct = CountsTable.from_tablename('example_table_tsv')
def test_initialize_fromdataset():
from singlet.counts_table import CountsTable
ct = CountsTable.from_datasetname('example_dataset')