def make_st_dataset(cnt_pths : List[str],
topn_genes : bool = None,
min_counts : int = 0,
min_spots : int = 0,
filter_genes : bool = False,
transpose : bool = False,
keep_barcodes: bool = True,
)-> CountData :
"""
Generate CountData object for ST-data
Parameter:
---------
cnt_pths : List[str]
list of paths to ST-data
topn_genes : bool
number of top expressed genes to
include in analysis
min_counts : int
minimal number of observed
counts assigned to a specific
spot/cell for it to be included
min_occurance : int
minimal number of occurances
of a gene among all spots/cells
for it to be included in analysis
filter_genes : bool
exclude MALAT1 and RP genes from
analysis
transpose : bool
transpose data
keep_barcodes: bool
if original rownames should be
used or whether rownames should
be based on spatial coordinates.
Only applicable when using h5ad files.
Returns:
-------
CountData object for the ST data
"""
# create joint matrix for count data
st_ext = utils.get_extenstion(cnt_pths[0])
if st_ext == "h5ad":
cnt = utils.read_h5ad_st(cnt_pths,keep_barcodes)
else:
cnt = utils.make_joint_matrix(cnt_pths,
transpose)
# select top N genes if specified
if topn_genes is not None:
genesize = cnt.values.sum(axis = 0)
topn_genes = np.min((topn_genes,genesize.shape[0]))
sel = np.argsort(genesize)[::-1]
sel = sel[0:topn_genes]
cnt = cnt.iloc[:,sel]
dataset = CountData(cnt)
# filter genes based on name
if filter_genes:
dataset.filter_genes()
# filter data based on quality
if any([min_counts > 0,min_spots > 0]):
dataset.filter_bad(min_counts = min_counts,
min_occurance = min_spots,
)
return dataset
def make_sc_dataset(cnt_pth : str,
lbl_pth : str,
topn_genes : int = None,
gene_list_pth : str = None,
filter_genes : bool = False,
lbl_colname : str = 'bio_celltype',
min_counts : int = 300,
min_cells : int = 0,
transpose : bool = False,
upper_bound : int = None,
lower_bound : int = None,
):
"""
Generate CountData object for SC-data
Parameter:
---------
cnt_pth : str
path to SC count data
lbl_pth : str
path to SC label data
topn_genes : bool
number of top expressed genes to
include
gene_list_pth : str
gene list
lbl_colname : str
name of column containing labels
min_counts : int
minimal number of observed
counts assigned to a specific
spot/cell for it to be included
min_cells : int
minimal number of occurances
of a gene among all cells
for it to be included
transpose : bool
transpose data
lower_bound : int
lower bound for the number of cells to
include from each type
upper_bound : int
upper bound for the number of cells to
include from each type
Returns:
-------
CountData object for the SC data
"""
sc_ext = utils.get_extenstion(cnt_pth)
if sc_ext == 'h5ad' :
cnt,lbl = utils.read_h5ad_sc(cnt_pth,
lbl_colname,
lbl_pth,
)
else:
cnt = utils.read_file(cnt_pth,sc_ext)
if transpose:
cnt = cnt.T
lbl = utils.read_file(lbl_pth)
# get labels
if lbl_colname is None:
lbl = lbl.iloc[:,0]
else:
lbl = lbl.loc[:,lbl_colname]
# match count and label data
inter = cnt.index.intersection(lbl.index)
if inter.shape[0] < 1:
print("[ERROR] : single cell count and annotation"\
" data did not match. Exiting.",
file = sys.stderr,
)
cnt = cnt.loc[inter,:]
lbl = lbl.loc[inter]
if upper_bound is not None or\
lower_bound is not None:
cnt,lbl = utils.subsample_data(cnt,
lbl,
lower_bound,
upper_bound,
)
# select top N expressed genes
if topn_genes is not None:
genesize = cnt.values.sum(axis = 0)
topn_genes = np.min((topn_genes,genesize.shape[0]))
sel = np.argsort(genesize)[::-1]
sel = sel[0:topn_genes]
cnt = cnt.iloc[:,sel]
# only use genes in specific genes list
# if specified
if gene_list_pth is not None:
with open(gene_list_pth,'r+') as fopen:
gene_list = fopen.readlines()
gene_list = pd.Index([ x.replace('\n','') for x in gene_list ])
sel = cnt.columns.intersection(gene_list)
cnt = cnt.loc[:,sel]
# create sc data set
dataset = CountData(cnt = cnt,
lbl = lbl)
# filter genes based on names
if filter_genes:
dataset.filter_genes()
# filter data based on quality
if any([min_counts > 0,min_cells > 0]):
dataset.filter_bad(min_counts = min_counts,
min_occurance = min_cells,
)
return dataset