def iterative_correction(matrix, args):
corrected_matrix, correction_factors = iterativeCorrection(
matrix, M=args.iterNum, verbose=args.verbose)
return corrected_matrix, correction_factors
def main(args=None):
args = parse_arguments().parse_args(args)
out = args.outPrefix
hic_matrix = HiCMatrix.hiCMatrix(matrixFile=args.matrixFile)
if args.overlapResolution is not None:
resolution = args.overlapResolution
else:
resolution = hic_matrix.getBinSize()
# filtering params
filterThreshold = list([-3, 3])
iterNum = 100
# compute spectra param
delta = 0.001
lookahead = 2
# normalize, call boundaries, and save as bed
for i in [10, 20, 30, 40, 50, 60, 70, 80, 90, 100]:
print("\n Sampling {}% of data \n".format(i))
# subsample
ma = copy.deepcopy(hic_matrix)
ma.matrix.data = ((hic_matrix.matrix.data) *
(float(i) / 100)).astype(int)
ma.matrix.eliminate_zeros()
# remove outliers
outlier_regions = hicCorrectMatrix.filter_by_zscore(
ma, filterThreshold[0], filterThreshold[1])
pct_outlier = 100 * float(len(outlier_regions)) / ma.matrix.shape[0]
ma.printchrtoremove(
outlier_regions,
label="Bins that are MAD outliers after merge ({:.2f}%) "
"out of".format(pct_outlier, ma.matrix.shape[0]))
# mask filtered regions
ma.maskBins(outlier_regions)
# total_filtered_out = set(outlier_regions)
# pre_row_sum = np.asarray(ma.matrix.sum(axis=1)).flatten()
# correct matrix
correction_factors = []
corrected_matrix, correction_factors = iterativeCorrection(
ma.matrix, iterNum)
ma.setMatrixValues(corrected_matrix)
ma.setCorrectionFactors(correction_factors)
# compute Spectra
chrom, chr_start, chr_end, matrix = hicFindTADs.compute_spectra_matrix(
args, matrix=ma)
# findTADs and save
min_idx = hicFindTADs.find_consensus_minima(matrix,
lookahead=lookahead,
delta=delta)
boundaries = np.array([chr_start[idx] for idx in min_idx])
chrom = chrom[min_idx]
bound_ext = np.array(
[boundaries - resolution, boundaries + resolution], dtype=int)
bound_ext = np.transpose(bound_ext)
outfile = "{}_{}.bed".format(out, i)
with open(outfile, 'w') as fh:
for i in range(bound_ext.shape[0]):
fh.write("{}\t{}\t{}\n".format(chrom[i], bound_ext[i, 0],
bound_ext[i, 1]))
# Intersect all the beds and plot
full = bt.bedtool.BedTool("{}_100.bed".format(out))
isect_all = np.array([])
for num in ['10', '20', '30', '40', '50', '60', '70', '80', '90', '100']:
bedfile = "{}_{}.bed".format(out, num)
bed = bt.bedtool.BedTool(bedfile)
isect = len(bed.intersect(full))
isect_all = np.append(isect_all, isect)
os.remove(bedfile)
isect_all = isect_all * 100 / isect_all[9]
x = np.arange(10., 110., 10.0)
plt.figure(figsize=(8, 8), dpi=300)
plt.plot(x, isect_all)
plt.xlabel('Sampling (%)')
plt.ylabel('TADs called (% w.r.t. total)')
plt.title('Overlap of TADs called per sample')
plt.savefig(out, dpi=300)