# flter for coverage
sys.stderr.write("Before cleaning there are " + str(len(hits.keys())) +
" hits\n")
hits = clean_hits()
sys.stderr.write("After cleaning there are " + str(len(hits.keys())) +
" hits\n")
# calculate the median for each position
sys.stderr.write("calculating medians\n")
median = []
mean = []
for i in range(longest_sequence):
data = [hits[s][i] for s in hits]
median.append(roblib.median(data))
mean.append(roblib.mean(data))
median = np.array(median)
mean = np.array(mean)
if subtract_medians:
for s in hits:
hits[s] = np.array(hits[s]) - median
if subtract_means:
for s in hits:
hits[s] = np.array(hits[s]) - mean
maxy = 0
for s in hits:
m = max(hits[s])
elif int(headers[i]) <= 63:
cols['control'].append(i)
else:
sys.stderr.write("Don't understand header {} at column {}\n".format(headers[i], i))
else:
data[p[0]] = map(int, p[1:])
allcontigs = data.keys()
allcontigs.sort()
# calculate the mean and stdev for each group and each contig
means = {}
std = {}
for contig in allcontigs:
means[contig] = {}
std[contig] = {}
for sample in cols:
testdata = [data[contig][i] for i in cols[sample]]
means[contig][sample] = roblib.mean(testdata)
std[contig][sample] = 2 * roblib.stdev(testdata)
# test the NS vs Control
if means[contig]['plasma'] - std[contig]['plasma'] > means[contig]['control'] + std[contig]['control'] and \
means[contig]['buffy'] - std[contig]['buffy'] > means[contig]['control'] + std[contig]['control'] and \
means[contig]['csf'] - std[contig]['csf'] > means[contig]['control'] + std[contig]['control']:
print("\t".join(map(str, ["ALL", contig, means[contig]['plasma'], means[contig]['buffy'], means[contig]['csf'],
means[contig]['control']])))
points = map(float, p[hcols:])
nz = filter(notzero, points)
psum = sum(points)
total += psum
data[p[0]] = [len(nz), psum]
# now calculate the mean and stdev based on the beta distribtion
xvalues = set()
betad = {}
for p in data:
if data[p][0] not in betad:
sys.stderr.write("alpha:" + str(data[p][0]) + " beta: " + str((ncols-data[p][0])+1)+ "\n")
# samples = np.random.beta(data[p][0]+1, ncols-data[p][0], 1000)
samples = np.random.beta(data[p][1]+1, total-data[p][1], 100000)
betad[data[p][0]] = (roblib.mean(samples), roblib.stdev(samples))
seen = set()
for p in data:
x = str(1.0 * data[p][0]/ncols)
sys.stdout.write(p + "\t" + str(x) + "\t" + str(1.0 * data[p][1]/total))
if x not in seen:
sys.stdout.write( "\t" + str(betad[data[p][0]][0]) + "\t" + str(betad[data[p][0]][1]))
seen.add(x)
sys.stdout.write("\n")
# flter for coverage
sys.stderr.write("Before cleaning there are " + str(len(hits.keys())) + " hits\n")
hits = clean_hits()
sys.stderr.write("After cleaning there are " + str(len(hits.keys())) + " hits\n")
# calculate the median for each position
sys.stderr.write("calculating medians\n")
median = []
mean = []
for i in range(longest_sequence):
data = [hits[s][i] for s in hits]
median.append(roblib.median(data))
mean.append(roblib.mean(data))
median = np.array(median)
mean = np.array(mean)
if subtract_medians:
for s in hits:
hits[s] = np.array(hits[s]) - median
if subtract_means:
for s in hits:
hits[s] = np.array(hits[s]) - mean
maxy = 0
for s in hits:
m = max(hits[s])
sys.stderr.write(
"Don't understand header {} at column {}\n".format(
headers[i], i))
else:
data[p[0]] = map(int, p[1:])
allcontigs = data.keys()
allcontigs.sort()
# calculate the mean and stdev for each group and each contig
means = {}
std = {}
for contig in allcontigs:
means[contig] = {}
std[contig] = {}
for sample in cols:
testdata = [data[contig][i] for i in cols[sample]]
means[contig][sample] = roblib.mean(testdata)
std[contig][sample] = 2 * roblib.stdev(testdata)
# test the NS vs Control
if means[contig]['plasma'] - std[contig]['plasma'] > means[contig]['control'] + std[contig]['control'] and \
means[contig]['buffy'] - std[contig]['buffy'] > means[contig]['control'] + std[contig]['control'] and \
means[contig]['csf'] - std[contig]['csf'] > means[contig]['control'] + std[contig]['control']:
print("\t".join(
map(str, [
"ALL", contig, means[contig]['plasma'], means[contig]['buffy'],
means[contig]['csf'], means[contig]['control']
])))
except:
sys.stderr.write("Can't add to position {}\n".format(
pu.reference_pos))
# here we trim the coverage array to make the math easier!
start = 1
end = args.l + 1
if args.e:
end = args.e + 1
coverage = coverage[0:end]
if args.s:
start = args.s
coverage = coverage[start:]
# calculate the average over coverage
av = mean(coverage)
st = stdev(coverage)
k = 0
for i, j in enumerate(coverage):
k += (j - av)**4
k = k / (len(coverage) * (st**4))
k -= 3
if args.r:
if args.c:
print("Filename\tReference\tAverage\tStDev\tKurtosis")
print(f"{args.f}\t{args.r}\t{av}\t{st}\t{k}")
else:
p = l.strip().split("\t")
if len(p) > ncols:
ncols = len(p)
points = map(float, p[hcols:])
nz = filter(notzero, points)
psum = sum(points)
total += psum
data[p[0]] = [len(nz), psum]
# now calculate the mean and stdev based on the beta distribtion
xvalues = set()
betad = {}
for p in data:
if data[p][0] not in betad:
sys.stderr.write("alpha:" + str(data[p][0]) + " beta: " +
str((ncols - data[p][0]) + 1) + "\n")
# samples = np.random.beta(data[p][0]+1, ncols-data[p][0], 1000)
samples = np.random.beta(data[p][1] + 1, total - data[p][1], 100000)
betad[data[p][0]] = (roblib.mean(samples), roblib.stdev(samples))
seen = set()
for p in data:
x = str(1.0 * data[p][0] / ncols)
sys.stdout.write(p + "\t" + str(x) + "\t" + str(1.0 * data[p][1] / total))
if x not in seen:
sys.stdout.write("\t" + str(betad[data[p][0]][0]) + "\t" +
str(betad[data[p][0]][1]))
seen.add(x)
sys.stdout.write("\n")
