def transfer_fields(segments, cnarr, ignore=params.IGNORE_GENE_NAMES):
"""Map gene names, weights, depths from `cnarr` bins to `segarr` segments.
Segment gene name is the comma-separated list of bin gene names. Segment
weight is the sum of bin weights, and depth is the (weighted) mean of bin
depths.
Also: Post-process segmentation output.
1. Ensure every chromosome has at least one segment.
2. Ensure first and last segment ends match 1st/last bin ends
(but keep log2 as-is).
"""
def make_null_segment(chrom, orig_start, orig_end):
"""Closes over 'segments'."""
vals = {'chromosome': chrom,
'start': orig_start,
'end': orig_end,
'gene': '-',
'depth': 0.0,
'log2': 0.0,
'probes': 0.0,
'weight': 0.0,
}
row_vals = tuple(vals[c] for c in segments.data.columns)
return row_vals
if not len(cnarr):
# This Should Never Happen (TM)
# raise RuntimeError("No bins for:\n" + str(segments.data))
logging.warn("No bins for:\n%s", segments.data)
return segments
# Adjust segment endpoints to cover the chromosome arm's original bins
# (Stretch first and last segment endpoints to match first/last bins)
bins_chrom = cnarr.chromosome.iat[0]
bins_start = cnarr.start.iat[0]
bins_end = cnarr.end.iat[-1]
if not len(segments):
# All bins in this chromosome arm were dropped: make a dummy segment
return make_null_segment(bins_chrom, bins_start, bins_end)
segments.start.iat[0] = bins_start
segments.end.iat[-1] = bins_end
# Aggregate segment depths, weights, gene names
# ENH refactor so that np/CNA.data access is encapsulated in skgenome
ignore += params.ANTITARGET_ALIASES
assert bins_chrom == segments.chromosome.iat[0]
cdata = cnarr.data.reset_index()
if 'depth' not in cdata.columns:
cdata['depth'] = np.exp2(cnarr['log2'].values)
bin_genes = cdata['gene'].values
bin_weights = cdata['weight'].values if 'weight' in cdata.columns else None
bin_depths = cdata['depth'].values
seg_genes = ['-'] * len(segments)
seg_weights = np.zeros(len(segments))
seg_depths = np.zeros(len(segments))
for i, bin_idx in enumerate(iter_slices(cdata, segments.data, 'outer', False)):
if bin_weights is not None:
seg_wt = bin_weights[bin_idx].sum()
if seg_wt > 0:
seg_dp = np.average(bin_depths[bin_idx],
weights=bin_weights[bin_idx])
else:
seg_dp = 0.0
else:
bin_count = len(cdata.iloc[bin_idx])
seg_wt = float(bin_count)
seg_dp = bin_depths[bin_idx].mean()
subgenes = [g for g in pd.unique(bin_genes[bin_idx]) if g not in ignore]
if subgenes:
seg_gn = ",".join(subgenes)
else:
seg_gn = '-'
seg_genes[i] = seg_gn
seg_weights[i] = seg_wt
seg_depths[i] = seg_dp
segments.data = segments.data.assign(
gene=seg_genes,
weight=seg_weights,
depth=seg_depths)
return segments
def transfer_fields(segments, cnarr, ignore=params.IGNORE_GENE_NAMES):
"""Map gene names, weights, depths from `cnarr` bins to `segarr` segments.
Segment gene name is the comma-separated list of bin gene names. Segment
weight is the sum of bin weights, and depth is the (weighted) mean of bin
depths.
Also: Post-process segmentation output.
1. Ensure every chromosome has at least one segment.
2. Ensure first and last segment ends match 1st/last bin ends
(but keep log2 as-is).
"""
def make_null_segment(chrom, orig_start, orig_end):
"""Closes over 'segments'."""
vals = {'chromosome': chrom,
'start': orig_start,
'end': orig_end,
'gene': '-',
'depth': 0.0,
'log2': 0.0,
'probes': 0.0,
'weight': 0.0,
}
row_vals = tuple(vals[c] for c in segments.data.columns)
return row_vals
if not len(cnarr):
# This Should Never Happen (TM)
# raise RuntimeError("No bins for:\n" + str(segments.data))
logging.warn("No bins for:\n%s", segments.data)
return segments
# Adjust segment endpoints to cover the chromosome arm's original bins
# (Stretch first and last segment endpoints to match first/last bins)
bins_chrom = cnarr.chromosome.iat[0]
bins_start = cnarr.start.iat[0]
bins_end = cnarr.end.iat[-1]
if not len(segments):
# All bins in this chromosome arm were dropped: make a dummy segment
return make_null_segment(bins_chrom, bins_start, bins_end)
segments.start.iat[0] = bins_start
segments.end.iat[-1] = bins_end
# Aggregate segment depths, weights, gene names
# ENH refactor so that np/CNA.data access is encapsulated in skgenome
ignore += params.ANTITARGET_ALIASES
assert bins_chrom == segments.chromosome.iat[0]
cdata = cnarr.data.reset_index()
if 'depth' not in cdata.columns:
cdata['depth'] = np.exp2(cnarr['log2'].values)
bin_genes = cdata['gene'].values
bin_weights = cdata['weight'].values if 'weight' in cdata.columns else None
bin_depths = cdata['depth'].values
seg_genes = ['-'] * len(segments)
seg_weights = np.zeros(len(segments))
seg_depths = np.zeros(len(segments))
for i, bin_idx in enumerate(iter_slices(cdata, segments.data, 'outer', False)):
if bin_weights is not None:
seg_wt = bin_weights[bin_idx].sum()
if seg_wt > 0:
seg_dp = np.average(bin_depths[bin_idx],
weights=bin_weights[bin_idx])
else:
seg_dp = 0.0
else:
bin_count = len(cdata.iloc[bin_idx])
seg_wt = float(bin_count)
seg_dp = bin_depths[bin_idx].mean()
subgenes = [g for g in pd.unique(bin_genes[bin_idx]) if g not in ignore]
if subgenes:
seg_gn = ",".join(subgenes)
else:
seg_gn = '-'
seg_genes[i] = seg_gn
seg_weights[i] = seg_wt
seg_depths[i] = seg_dp
segments.data = segments.data.assign(
gene=seg_genes,
weight=seg_weights,
depth=seg_depths)
return segments