def concat_data(self, modality: str = "rna"):
""" Used for raw data, Ignore multiarrays and only consider one matrix per unidata """
genomes = []
unidata_arr = []
for key in list(self.data):
unidata = self.data.pop(key)
assert unidata.get_modality() == modality
genomes.append(unidata.get_genome())
unidata_arr.append(unidata)
unikey = None
if len(genomes) == 1:
unikey = unidata_arr[0].get_uid()
self.data[unikey] = unidata_arr[0]
else:
genome = ",".join(genomes)
feature_metadata = pd.concat(
[unidata.feature_metadata for unidata in unidata_arr], axis=0)
feature_metadata.reset_index(inplace=True)
feature_metadata.fillna(value="N/A", inplace=True)
X = hstack([unidata.matrices["X"] for unidata in unidata_arr],
format="csr")
unidata = UnimodalData(unidata_arr[0].barcode_metadata,
feature_metadata, {"X": X}, {
"genome": genome,
"modality": "rna"
})
unikey = unidata.get_uid()
self.data[unikey] = unidata
del unidata_arr
gc.collect()
self._selected = unikey
self._unidata = self.data[unikey]
def _generate_filter_plots(unidata: UnimodalData,
plot_filt: str,
plot_filt_figsize: str = None,
min_genes_before_filt: int = 100) -> None:
""" This function generates filtration plots, only used in command line.
"""
group_key = unidata.get_uid()
from pegasus.plotting import qcviolin
kwargs = {"return_fig": True, "dpi": 500}
if plot_filt_figsize is not None:
width, height = plot_filt_figsize.split(",")
kwargs["panel_size"] = (int(width), int(height))
fig = qcviolin(unidata, "count", **kwargs)
fig.savefig(f"{plot_filt}.{group_key}.filt.UMI.pdf")
fig = qcviolin(unidata, "gene", **kwargs)
fig.savefig(f"{plot_filt}.{group_key}.filt.gene.pdf")
fig = qcviolin(unidata, "mito", **kwargs)
if fig is not None:
fig.savefig(f"{plot_filt}.{group_key}.filt.mito.pdf")
logger.info("Filtration plots are generated.")
def add_data(self, unidata: UnimodalData) -> None:
""" Add data, if _selected is not set, set as the first added dataset
"""
key = unidata.get_uid()
assert key is not None
if key in self.data:
raise ValueError(f"Key '{key}' already exists!")
self.data[key] = unidata
if self._selected is None:
self._selected = key
self._unidata = unidata