def test_exclude_bad_probes(self):
logging.info("Testing removing bad probes...")
probes_meth_data = self.meth_data.copy()
data_no_bad_probes = methylation_data.MethylationDataLoader(
datafile=self.DEMO_DATA_NO_BAD_PROBES)
bad_probes = load(open(self.BAD_PROBES, 'r'))
module = refactor.Refactor(methylation_data=probes_meth_data,
k=5,
bad_probes_list=bad_probes)
module._exclude_bad_probes()
assert array_equal(data_no_bad_probes.data, module.meth_data.data)
# tests sites list has changed
remove_count = len(bad_probes)
orig_sites_before = []
orig_sites_before.extend(self.meth_data.cpgnames)
orig_sites_after = []
orig_sites_after.extend(module.meth_data.cpgnames)
for i in bad_probes:
try:
orig_sites_before.remove(i)
except:
remove_count -= 1
assert orig_sites_after == orig_sites_before
# test sites size
assert self.meth_data.sites_size - remove_count == module.meth_data.sites_size
logging.info("PASS")
def test_low_rank_approx_distances(self):
"""
tests that number of distances is as the number of sites (distance for every site)
"""
logging.info("Testing low rank approx distances...")
dis_meth_data = self.meth_data.copy()
module = refactor.Refactor(methylation_data=dis_meth_data, k=5)
distances = module._calc_low_rank_approx_distances(dis_meth_data)
assert distances.size == dis_meth_data.sites_size, "there must be distances as the number of sites"
logging.info("PASS")
def test_senario1(self):
logging.info("Testing clean refactor components...")
refactor_meth_data = self.meth_data.copy()
comp = loadtxt(self.RES1)
ranked = loadtxt(self.RES1_RANKED, dtype=str)
module = refactor.Refactor(methylation_data=refactor_meth_data,
k=5,
t=500,
stdth=0,
bad_probes_list=self.bad_probes,
use_phenos=None,
use_covars=None)
module.run()
assert self._validate_ranked_lists(module, ranked)
assert module.components.shape == comp.shape
for i in range(module.components.shape[1]):
assert tools.correlation(module.components[:, i], comp[:, i])
logging.info("PASS")
def test_senario3(self):
logging.info("Testing controls feature selection...")
refactor_meth_data = self.meth_data.copy()
comp = loadtxt(self.RES3)
ranked = loadtxt(self.RES3_RANKED, dtype=str)
module = refactor.Refactor(methylation_data=refactor_meth_data,
k=5,
t=500,
stdth=0,
bad_probes_list=self.bad_probes,
feature_selection='controls',
use_phenos=['p2'],
use_covars=[])
module.run()
assert self._validate_ranked_lists(module, ranked)
assert module.components.shape == comp.shape
for i in range(module.components.shape[1]):
assert tools.correlation(module.components[:, i], comp[:, i])
logging.info("PASS")
def run(self, args, meth_data, output_perfix = None):
try:
if args.pheno is not None and meth_data.phenotype is None:
common.terminate("There is no phenotype in the data, use --phenofile to add phenotype.")
if not output_perfix:
output_perfix = "output"
bad_probes_list = set()
[bad_probes_list.update(loadtxt(probes_file, dtype=str)) for probes_file in BAD_PROBES_FILES]
bad_probes_list = array(list(bad_probes_list))
self.module = refactor.Refactor(methylation_data = meth_data,
k = args.k,
t = args.t,
stdth = args.stdth,
feature_selection = args.fs.lower().strip(),
num_components = args.numcomp,
use_covars = args.covar,
use_phenos = args.pheno,
bad_probes_list = bad_probes_list,
ranked_output_filename = output_perfix + "." + refactor.RANKED_FILENAME,
components_output_filename = output_perfix + "." + refactor.COMPONENTS_FILENAME)
self.module.run()
except Exception :
logging.exception("in refactor")
raise