def load_epi_TMS(average='geometric'):
from singlet import Dataset
data_fn = '../../data/tabulamurissenis/lung_epi_droplet.h5ad'
dsepi = Dataset(
dataset={
'path': data_fn,
'index_samples': 'obs_names',
'index_features': 'var_names',
'bit_precision': 32,
})
dsepi.samplesheet['Timepoint'] = 'Adult'
dsepi.samplesheet['cellSubtype'] = dsepi.samplesheet[
'free_annotation'].replace({
'Alveolar Epithelial Type 2': 'AT2',
'Alveolar Epithelial Type 1': 'AT1',
})
dsepi.counts.normalize(inplace=True)
if not average:
return dsepi
if average == 'geometric':
dsepi.counts.log(inplace=True)
dsepi_av = dsepi.average('samples', by=['cellSubtype', 'Timepoint'])
return dsepi_av
def load_epi_development(average='geometric'):
from singlet import Dataset
data_fn = '../../data/Cohen_et_al_2018/cohen2018_lung_epi_cpm.loom'
dsepi = Dataset(
dataset={
'path': data_fn,
'index_samples': 'obs_names',
'index_features': 'var_names',
'bit_precision': 32,
})
dsepi.counts.normalize(inplace=True)
if not average:
return dsepi
if average == 'geometric':
dsepi.counts.log(inplace=True)
dsepi_av = dsepi.average('samples', by=['cellSubtype', 'Timepoint'])
return dsepi_av
fdn_data = '../../data/sequencing/datasets/endo_final/'
fns_loom = {
'ours': fdn_data + 'endo_normoxia.loom',
'tabulamuris': '../../data/tabulamuris/FACS_alltissues/endos.loom',
}
fig_fdn = '../../figures/endomese_share/endo_paper_figure_2/'
if __name__ == '__main__':
os.makedirs(fig_fdn, exist_ok=True)
print('Load our endo data')
ds = Dataset(dataset={
'path': fns_loom['ours'],
'index_samples': '_index',
'index_features': 'GeneName',
}, )
ds.samplesheet['Tissue'] = 'Lung'
ds.samplesheet.index.name = 'CellID'
ds.counts.columns.name = 'CellID'
ds.samplesheet['Dataset'] = 'ours'
print('Load Tabula Muris FACS endothelial')
ds_tm = Dataset(dataset={
'path': fns_loom['tabulamuris'],
'index_samples': 'CellID',
'index_features': 'GeneName',
}, )
ds_tm.samplesheet['Tissue'] = ds_tm.samplesheet['tissue']
ds_tm.samplesheet['Mousename'] = 'TM' + ds_tm.samplesheet['mouse.id']
)
if __name__ == '__main__':
if not os.path.isfile(fns_loom['ours']):
create_loom_from_figshare(fns_loom['ours'])
if not os.path.isfile(fns_loom['tabulamuris']):
ss = create_loom_from_tabulamurisfacs(fns_loom['tabulamuris'])
print('Load loom file')
ds = Dataset(
dataset={
'path': fns_loom['ours'],
'index_samples': 'CellID',
'index_features': 'GeneName',
},
)
ds.samplesheet['cellSubtype'] = ds.samplesheet['Cell Subtype']
ds.query_samples_by_metadata(
'cellSubtype in ("Mac I", "Mac II", "Mac III", "Mac IV", "Mac V")',
inplace=True)
print('Load Schyns et al loom file')
ds_sc = Dataset(
dataset={
'path': fns_loom['schyns'],
'index_samples': 'CellID',
'index_features': 'Gene',
'bit_precision': 32,
'pancreas',
'skin',
'spleen',
'thymus',
'tongue',
'trachea',
]
if __name__ == '__main__':
pa = argparse.ArgumentParser(
description='Test loading various tabula muris tissues', )
pa.add_argument(
'--tissues',
default=tissues,
nargs='+',
choices=tissues,
help='Tissue of origin',
)
args = pa.parse_args()
dss = {}
for tissue in args.tissues:
print(tissue)
ds = Dataset(
counts_table=tissue,
samplesheet='alltissues',
featuresheet='alltissues',
)
dss[tissue] = ds
print('Parse gene metadata')
fn_pbmc_meta_genes = '../data/pbmc_zanini/featuresheet_10_10_unique_L1.tsv'
featuresheet = pd.read_csv(fn_pbmc_meta_genes, sep='\t', index_col=0)
c = Counter(featuresheet['GeneName'])
ind = []
for fea in featuresheet.index:
if c[featuresheet.at[fea, 'GeneName']] == 1:
ind.append(fea)
featuresheet = featuresheet.loc[ind]
counts = counts.loc[ind]
print('Reannotate cell types')
ds = Dataset(
samplesheet=cells,
counts_table=counts,
featuresheet=featuresheet,
)
ds.reindex(axis='features', column='GeneName', inplace=True)
print('Filter low-quality cells')
ds.samplesheet['coverage'] = ds.samplesheet['coverage'].astype(int)
ds.query_samples_by_metadata('coverage >= 50000', inplace=True)
print('Ignore HLA types and TCR and BCR variable regions')
def fun(fn):
if fn in ('HLA-A', 'HLA-B', 'HLA-C'):
return False
if fn.startswith('TRBV'):
return False
n_genes = (counts_table.iloc[len(counts_table._spikeins
):-len(counts_table._otherfeatures)]
>= 1).sum(axis=0)
samplesheet['number_of_genes_1plusreads'] = n_genes
print('Exclude slashes for loom')
featuresheet.rename(columns={'Chromosome/scaffold name': 'Chromosome'},
inplace=True)
print('Add index name for loom')
samplesheet.index.name = 'CellID'
counts_table.columns.name = 'CellID'
ds = Dataset(
counts_table=counts_table,
featuresheet=featuresheet,
samplesheet=samplesheet,
)
print('Save raw dataset as loom file')
fn_raw = '../../data/sequencing/datasets/all_{:}/raw.loom'.format(
version)
if not args.dry:
ds.to_dataset_file(fn_raw, fmt='loom')
# BLOCK 2: filter genes and cells
if args.block == 2:
print('Load raw loom file')
import loompy
fn_raw = '../../data/sequencing/datasets/all_{:}/raw.loom'.format(
print('Feature selection')
features = ds.feature_selection.overdispersed_within_groups('Mousename',
inplace=False)
dsf = ds.query_features_by_name(features)
print('PCA')
dsc = dsf.dimensionality.pca(n_dims=30,
robust=False,
return_dataset='samples')
print('Smoothen along knn')
edges = dsc.graph.knn('samples', 20, return_kind='edges')
# Smoothen twice (reaching second-order neighbors)
counts_smooth = ds.counts.smoothen_neighbors(edges, n_iterations=2)
ds_smooth = Dataset(
counts_table=counts_smooth,
samplesheet=ds.samplesheet.copy(),
)
print('Plot known signaling genes, smoothened')
genes = [g for g in genes_all if (g in ds.featurenames)]
markers = list(genes) + [
'cellSubtype',
]
fig, axs = plt.subplots(4, 5, figsize=(12, 8), sharex=True, sharey=True)
axs = axs.ravel()
for i, (gene, ax) in enumerate(zip(markers, axs)):
if gene == 'cellSubtype':
cmap = {
'Mac IV': 'darkolivegreen',
'Mac II': 'lime',
row_attrs=row_attrs,
)
if __name__ == '__main__':
if not os.path.isfile(fns_loom['ours']):
create_loom_from_figshare(fns_loom['ours'])
if not os.path.isfile(fns_loom['tabulamuris']):
ss = create_loom_from_tabulamurisfacs(fns_loom['tabulamuris'])
print('Load loom file')
ds = Dataset(dataset={
'path': fns_loom['ours'],
'index_samples': 'CellID',
'index_features': 'GeneName',
}, )
ds.samplesheet['Tissue'] = 'lung'
print('Load Tabula Muris FACS immune')
ds_tm = Dataset(dataset={
'path': fns_loom['tabulamuris'],
'index_samples': 'CellID',
'index_features': 'GeneName',
}, )
ds_tm.samplesheet['Tissue'] = ds_tm.samplesheet['tissue']
ds_tm.samplesheet['Cell Subtype'] = ds_tm.samplesheet[
'cell_ontology_class']
ds_tm.samplesheet['Cell Subtype'].replace(
{