def convert_files():
print 'Converting files to CSV...'
filenames = map(utils.datafile, get_filenames())
convert(filenames[0], filenames[1], utils.datafile('mnist-train.csv'), 60000)
convert(filenames[2], filenames[3], utils.datafile('mnist-test.csv'), 10000)
for fn in filenames:
os.remove(fn)
def download_files():
base_url = 'http://yann.lecun.com/exdb/mnist/'
for fn in get_filenames():
file_url = base_url+fn+'.gz'
local_path = utils.datafile(fn+'.gz')
print 'Saving {0} to {1}'.format(file_url, local_path)
urllib.urlretrieve(file_url,local_path)
def ensure_datadir_exists():
datadir = utils.datafile('')
if os.path.exists(datadir):
print 'Directory {0} already exists; skipping'.format(datadir)
else:
print 'Creating directory {0}'.format(datadir)
os.makedirs(datadir)
def extract_files():
for fn in map(utils.datafile, get_filenames()):
with gzip.open(fn+'.gz', 'rb') as gzf:
with open(utils.datafile(fn), 'wb') as f:
print 'Extracting {0}'.format(gzf.name)
f.write(gzf.read())
print 'Deleting {0}.gz'.format(fn)
os.remove(fn+'.gz')
def get_Y_depth(self, N=5):
UNIQY = datafile("chrY.{}.unique_ccn.gc".format(
self.ref.split('_')[0]))
fp = open(UNIQY)
depths = []
for i, row in enumerate(fp):
# Some regions still have mapped reads, exclude a few
if i in (1, 4, 6, 7, 10, 11, 13, 16, 18, 19):
continue
if len(depths) >= N:
break
c, start, end, gc = row.split()
start, end = int(start), int(end)
d = self.region_depth(c, start, end)
depths.append(d)
self.logger.debug("Y depths (first {} regions): {}"\
.format(N, np.array(depths)))
return np.median(depths)
import numpy as np
from math import exp
from collections import defaultdict
from bam_parser import PEextractor, FLANKMATCH, SPAN
from utils import datafile, listify
from scipy.stats import gaussian_kde, poisson
# Global settings
MAX_PERIOD = 6
SMALL_VALUE = exp(-10)
REALLY_SMALL_VALUE = exp(-100)
MODEL_PREFIX = "illumina_v3.pcrfree"
STEPMODEL = datafile(MODEL_PREFIX + ".stepmodel")
NOISEMODEL = datafile(MODEL_PREFIX + ".stuttermodel")
MIN_SPANNING_PAIRS = 5
class StepModel:
"""
Contains information about step size distributions
"""
def __init__(self, filename=STEPMODEL):
self.non_unit_step_by_period = {}
self.step_size_by_period = {}
fp = open(filename, "r")
for i in range(MAX_PERIOD):
line = fp.readline()
self.non_unit_step_by_period[i + 1] = float(line.strip())
