def mergecn(args):
"""
%prog mergecn FACE.csv
Compile matrix of GC-corrected copy numbers. Place a bunch of folders in
csv file. Each folder will be scanned, one chromosomes after another.
"""
p = OptionParser(mergecn.__doc__)
opts, args = p.parse_args(args)
if len(args) != 1:
sys.exit(not p.print_help())
(csvfile, ) = args
samples = [x.replace("-cn", "").strip().strip("/") for x in open(csvfile)]
betadir = "beta"
mkdir(betadir)
for seqid in allsomes:
names = [
op.join(s + "-cn", "{}.{}.cn".format(op.basename(s), seqid))
for s in samples
]
arrays = [np.fromfile(name, dtype=np.float) for name in names]
shapes = [x.shape[0] for x in arrays]
med_shape = np.median(shapes)
arrays = [x for x in arrays if x.shape[0] == med_shape]
ploidy = 2 if seqid not in ("chrY", "chrM") else 1
if seqid in sexsomes:
chr_med = [np.median([x for x in a if x > 0]) for a in arrays]
chr_med = np.array(chr_med)
idx = get_kmeans(chr_med, k=2)
zero_med = np.median(chr_med[idx == 0])
one_med = np.median(chr_med[idx == 1])
logging.debug("K-means with {} c0:{} c1:{}".format(
seqid, zero_med, one_med))
higher_idx = 1 if one_med > zero_med else 0
# Use the higher mean coverage componen
arrays = np.array(arrays)[idx == higher_idx]
arrays = [[x] for x in arrays]
ar = np.concatenate(arrays)
print(seqid, ar.shape)
rows, columns = ar.shape
beta = []
std = []
for j in range(columns):
a = ar[:, j]
beta.append(np.median(a))
std.append(np.std(a) / np.mean(a))
beta = np.array(beta) / ploidy
betafile = op.join(betadir, "{}.beta".format(seqid))
beta.tofile(betafile)
stdfile = op.join(betadir, "{}.std".format(seqid))
std = np.array(std)
std.tofile(stdfile)
logging.debug("Written to `{}`".format(betafile))
ar.tofile("{}.bin".format(seqid))
def mergecn(args):
"""
%prog mergecn FACE.csv
Compile matrix of GC-corrected copy numbers. Place a bunch of folders in
csv file. Each folder will be scanned, one chromosomes after another.
"""
p = OptionParser(mergecn.__doc__)
opts, args = p.parse_args(args)
if len(args) != 1:
sys.exit(not p.print_help())
csvfile, = args
samples = [x.replace("-cn", "").strip().strip("/") for x in open(csvfile)]
betadir = "beta"
mkdir(betadir)
for seqid in allsomes:
names = [op.join(s + "-cn", "{}.{}.cn".
format(op.basename(s), seqid)) for s in samples]
arrays = [np.fromfile(name, dtype=np.float) for name in names]
shapes = [x.shape[0] for x in arrays]
med_shape = np.median(shapes)
arrays = [x for x in arrays if x.shape[0] == med_shape]
ploidy = 2 if seqid not in ("chrY", "chrM") else 1
if seqid in sexsomes:
chr_med = [np.median([x for x in a if x > 0]) for a in arrays]
chr_med = np.array(chr_med)
idx = get_kmeans(chr_med, k=2)
zero_med = np.median(chr_med[idx == 0])
one_med = np.median(chr_med[idx == 1])
logging.debug("K-means with {} c0:{} c1:{}"
.format(seqid, zero_med, one_med))
higher_idx = 1 if one_med > zero_med else 0
# Use the higher mean coverage componen
arrays = np.array(arrays)[idx == higher_idx]
arrays = [[x] for x in arrays]
ar = np.concatenate(arrays)
print seqid, ar.shape
rows, columns = ar.shape
beta = []
std = []
for j in xrange(columns):
a = ar[:, j]
beta.append(np.median(a))
std.append(np.std(a) / np.mean(a))
beta = np.array(beta) / ploidy
betafile = op.join(betadir, "{}.beta".format(seqid))
beta.tofile(betafile)
stdfile = op.join(betadir, "{}.std".format(seqid))
std = np.array(std)
std.tofile(stdfile)
logging.debug("Written to `{}`".format(betafile))
ar.tofile("{}.bin".format(seqid))
def calibrate(args):
"""
%prog calibrate calibrate.JPG boxsize
Calibrate pixel-inch ratio and color adjustment.
- `calibrate.JPG` is the photo containig a colorchecker
- `boxsize` is the measured size for the boxes on printed colorchecker, in
squared centimeter (cm2) units
"""
xargs = args[2:]
p = OptionParser(calibrate.__doc__)
opts, args, iopts = add_seeds_options(p, args)
if len(args) != 2:
sys.exit(not p.print_help())
imagefile, boxsize = args
boxsize = float(boxsize)
# Read in color checker
colorcheckerfile = op.join(datadir, "colorchecker.txt")
colorchecker = []
expected = 0
for row in open(colorcheckerfile):
boxes = row.split()
colorchecker.append(boxes)
expected += len(boxes)
folder = op.split(imagefile)[0]
objects = seeds([imagefile, "--outdir={0}".format(folder)] + xargs)
nseeds = len(objects)
logging.debug("Found {0} boxes (expected={1})".format(nseeds, expected))
assert expected - 4 <= nseeds <= expected + 4, \
"Number of boxes drastically different from {0}".format(expected)
# Calculate pixel-cm ratio
boxes = [t.area for t in objects]
reject = reject_outliers(boxes)
retained_boxes = [b for r, b in zip(reject, boxes) if not r]
mbox = np.median(retained_boxes) # in pixels
pixel_cm_ratio = (mbox / boxsize) ** .5
logging.debug("Median box size: {0} pixels. Measured box size: {1} cm2".\
format(mbox, boxsize))
logging.debug("Pixel-cm ratio: {0}".format(pixel_cm_ratio))
xs = [t.x for t in objects]
ys = [t.y for t in objects]
idx_xs = get_kmeans(xs, 6)
idx_ys = get_kmeans(ys, 4)
for xi, yi, s in zip(idx_xs, idx_ys, objects):
s.rank = (yi, xi)
objects.sort(key=lambda x: x.rank)
colormap = []
for s in objects:
x, y = s.rank
observed, expected = s.rgb, rgb_to_triplet(colorchecker[x][y])
colormap.append((np.array(observed), np.array(expected)))
# Color transfer
tr0 = np.eye(3).flatten()
print("Initial distance:", total_error(tr0, colormap), file=sys.stderr)
tr = fmin(total_error, tr0, args=(colormap,))
tr.resize((3, 3))
print("RGB linear transform:\n", tr, file=sys.stderr)
calib = {"PixelCMratio": pixel_cm_ratio,
"RGBtransform": tr.tolist()}
jsonfile = op.join(folder, "calibrate.json")
fw = must_open(jsonfile, "w")
print(json.dumps(calib, indent=4), file=fw)
fw.close()
logging.debug("Calibration specs written to `{0}`.".format(jsonfile))
return jsonfile
def calibrate(args):
"""
%prog calibrate calibrate.JPG boxsize
Calibrate pixel-inch ratio and color adjustment.
- `calibrate.JPG` is the photo containig a colorchecker
- `boxsize` is the measured size for the boxes on printed colorchecker, in
squared centimeter (cm2) units
"""
xargs = args[2:]
p = OptionParser(calibrate.__doc__)
opts, args, iopts = add_seeds_options(p, args)
if len(args) != 2:
sys.exit(not p.print_help())
imagefile, boxsize = args
boxsize = float(boxsize)
# Read in color checker
colorcheckerfile = op.join(datadir, "colorchecker.txt")
colorchecker = []
expected = 0
for row in open(colorcheckerfile):
boxes = row.split()
colorchecker.append(boxes)
expected += len(boxes)
folder = op.split(imagefile)[0]
objects = seeds([imagefile, "--outdir={0}".format(folder)] + xargs)
nseeds = len(objects)
logging.debug("Found {0} boxes (expected={1})".format(nseeds, expected))
assert (
expected - 4 <= nseeds <= expected +
4), "Number of boxes drastically different from {0}".format(expected)
# Calculate pixel-cm ratio
boxes = [t.area for t in objects]
reject = reject_outliers(boxes)
retained_boxes = [b for r, b in zip(reject, boxes) if not r]
mbox = np.median(retained_boxes) # in pixels
pixel_cm_ratio = (mbox / boxsize)**0.5
logging.debug(
"Median box size: {0} pixels. Measured box size: {1} cm2".format(
mbox, boxsize))
logging.debug("Pixel-cm ratio: {0}".format(pixel_cm_ratio))
xs = [t.x for t in objects]
ys = [t.y for t in objects]
xs = [float(itemx) for itemx in xs]
ys = [float(itemy) for itemy in ys]
idx_xs = get_kmeans(xs, 6)
idx_ys = get_kmeans(ys, 4)
for xi, yi, s in zip(idx_xs, idx_ys, objects):
s.rank = (yi, xi)
objects.sort(key=lambda x: x.rank)
colormap = []
for s in objects:
x, y = s.rank
observed, expected = s.rgb, rgb_to_triplet(colorchecker[x][y])
colormap.append((np.array(observed), np.array(expected)))
# Color transfer
tr0 = np.eye(3).flatten()
print("Initial distance:", total_error(tr0, colormap), file=sys.stderr)
tr = fmin(total_error, tr0, args=(colormap, ))
tr.resize((3, 3))
print("RGB linear transform:\n", tr, file=sys.stderr)
calib = {"PixelCMratio": pixel_cm_ratio, "RGBtransform": tr.tolist()}
jsonfile = op.join(folder, "calibrate.json")
fw = must_open(jsonfile, "w")
print(json.dumps(calib, indent=4), file=fw)
fw.close()
logging.debug("Calibration specs written to `{0}`.".format(jsonfile))
return jsonfile
