def test_bigwigwriter_not_sorted(tmpdir):
from kipoi.writers import BigWigWriter
import pyBigWig
import tempfile
tmpfile = str(tmpdir.mkdir("example").join("out.bw"))
bww = BigWigWriter(tmpfile, chrom_sizes=[('chr1', 1000), ('chr10', 1000)], is_sorted=False)
regions = [
({"chr": "chr1", "start": 30, "end": 40}, np.arange(10)[::-1]),
({"chr": "chr1", "start": 10, "end": 20}, np.arange(10)),
({"chr": "chr10", "start": 10, "end": 20}, np.arange(10))
]
for region, data in regions:
bww.region_write(region, data)
bww.close()
# load the bigwig file and validate the values
r = pyBigWig.open(tmpfile)
for region, data in regions:
# query the values
assert np.allclose(data, r.values(region['chr'],
region['start'],
region['end'], numpy=True))
# assert there are no values here
assert np.isnan(r.values("chr1", 20, 30, numpy=True)).all()
r.close()
def test_bigwigwriter():
from kipoi.writers import BigWigWriter
import pyBigWig
import tempfile
temp_path = tempfile.mkstemp()[1]
with pytest.raises(Exception):
bww = BigWigWriter(temp_path)
regions = {"chr": ["chr1", "chr7", "chr2"], "start": [10, 30, 20], "end": [11, 31, 21]}
values = [3.0, 4.0, 45.4]
for i, val in enumerate(values):
reg = {k: v[i] for k, v in regions.items()}
bww.region_write(reg, np.array([val]))
bww.close()
bww_2 = pyBigWig(temp_path)
for i, val in enumerate(values):
reg = {k: v[i] for k, v in regions.items()}
bww.region_write(reg, [val])
assert bww_2.entries(reg["chr"], reg["start"], reg["end"])[0][2] == val
def contrib2bw(contrib_file, output_prefix):
"""Convert the contribution file to bigwigs
"""
from kipoi.writers import BigWigWriter
from bpnet.cli.contrib import ContribFile
from bpnet.cli.modisco import get_nonredundant_example_idx
output_dir = os.path.dirname(output_prefix)
add_file_logging(output_dir, logger, 'contrib2bw')
os.makedirs(output_dir, exist_ok=True)
cf = ContribFile(contrib_file)
# remove overlapping intervals
ranges = cf.get_ranges()
keep_idx = get_nonredundant_example_idx(ranges, width=None)
cf.include_samples = keep_idx
discarded = len(ranges) - len(keep_idx)
logger.info(
f"{discarded}/{len(ranges)} of ranges will be discarded due to overlapping intervals"
)
contrib_scores = cf.available_contrib_scores(
) # TODO - implement contrib_wildcard to filter them
chrom_sizes = [(k, v) for k, v in cf.get_chrom_sizes().items()]
ranges = cf.ranges()
assert len(ranges) == len(keep_idx)
delim = "." if not output_prefix.endswith("/") else ""
for contrib_score in contrib_scores:
contrib_dict = cf.get_contrib(contrib_score=contrib_score)
contrib_score_name = contrib_score.replace("/", "_")
for task, contrib in contrib_dict.items():
output_file = output_prefix + f'{delim}.contrib.{contrib_score_name}.{task}.bw'
logger.info(f"Genrating {output_file}")
contrib_writer = BigWigWriter(output_file,
chrom_sizes=chrom_sizes,
is_sorted=False)
for idx in range(len(ranges)):
contrib_writer.region_write(region={
"chr": ranges['chrom'].iloc[idx],
"start": ranges['start'].iloc[idx],
"end": ranges['end'].iloc[idx]
},
data=contrib[idx])
contrib_writer.close()
logger.info("Done!")