def _create_query_dmr_ds(reptile, dmr_regions_bed_df):
query_dmr_ds = RegionDS.from_bed(
dmr_regions_bed_df,
chrom_size_path=reptile.chrom_size_path,
location=reptile.output_path,
region_dim="query-dmr",
)
query_dmr_ds.save()
return
def _create_query_region_ds(reptile):
pybedtools.BedTool().makewindows(
g=reptile.chrom_size_path, s=reptile.step_size, w=reptile.window_size
).saveas(f"{reptile.output_path}/query_region.bed")
query_region_ds = RegionDS.from_bed(
f"{reptile.output_path}/query_region.bed",
chrom_size_path=reptile.chrom_size_path,
location=reptile.output_path,
region_dim="query-region",
)
subprocess.run(f"rm -f {reptile.output_path}/query_region.bed", shell=True)
query_region_ds.save()
return
def _create_train_dmr_ds(reptile, train_regions_bed, train_label):
# total DMRs
dmr_regions_bed = pybedtools.BedTool(reptile.dmr_regions).sort(
g=reptile.chrom_size_path
)
dmr_regions_bed_df = dmr_regions_bed.to_dataframe()
# train DMRs and train DMR labels
train_dmr = train_regions_bed.map(dmr_regions_bed, c=4, o="collapse").to_dataframe()
dmr_label = defaultdict(list)
for _, row in train_dmr.iterrows():
*_, train_region, dmrs = row
if dmrs == ".":
continue
dmrs = dmrs.split(",")
for dmr in dmrs:
dmr_label[dmr].append(train_label[train_region])
# some DMR might have multiple labels
consistent_dmr_label = {}
for dmr, dmr_labels in dmr_label.items():
if (len(dmr_labels) == 1) or (len(set(dmr_labels)) == 1):
consistent_dmr_label[dmr] = dmr_labels[0]
else:
# dmr has in consistent label
continue
dmr_label = pd.Series(consistent_dmr_label)
dmr_label.index.name = "train-dmr"
train_dmr_regions_bed_df = (
dmr_regions_bed_df.set_index("name")
.loc[dmr_label.index]
.reset_index()
.iloc[:, [1, 2, 3, 0]]
)
# train DMR RegionDS
train_dmr_ds = RegionDS.from_bed(
train_dmr_regions_bed_df,
chrom_size_path=reptile.chrom_size_path,
location=reptile.output_path,
region_dim="train-dmr",
)
train_dmr_ds.coords["train-dmr_label"] = dmr_label
train_dmr_ds.save()
return dmr_regions_bed_df
def _create_train_region_ds(reptile):
# train regions
train_regions_bed = pybedtools.BedTool(reptile.train_regions).sort(
g=reptile.chrom_size_path
)
# train region labels
train_label = pd.read_csv(
reptile.train_region_labels, sep="\t", index_col=0, squeeze=True
)
train_label.index.name = "train-region"
train_regions_bed_df = train_regions_bed.to_dataframe()
# train RegionDS
train_region_ds = RegionDS.from_bed(
train_regions_bed_df,
chrom_size_path=reptile.chrom_size_path,
location=reptile.output_path,
region_dim="train-region",
)
train_region_ds.coords["train-region_label"] = train_label
train_region_ds.save()
return train_regions_bed, train_label