def diploid_sample_from_labels(ls=None,
ref=None,
hp1=None,
hp2=None):
"""Create `medaka.common.Sample` objects from a specified
reference string and haplotype strings for easily
mocking variant calling scenarios.
"""
assert len(ref) == len(hp1) == len(hp2)
pos = mock_positions_array(ref)
# mocking up the network output in terms of 0. and 1.
# in reality they would be float probabilities
probs = np.zeros((len(pos), len(ls._decoding)))
for i in range(len(ref)):
diploid_label = tuple(sorted((hp1[i], hp2[i])))
probs[i, ls._encoding[diploid_label]] = 1
s = common.Sample(ref_name='contig1', features=None,
labels=None, ref_seq=None,
positions=pos, label_probs=probs)
return s, ref
def test_decode_consensus(self):
"""Test the conversion between network outputs and sequence"""
num_classes = 13 # 3 elements per base * 4 bases + *
label_probs = np.zeros([6, num_classes])
label_probs[0, 10] = 0.9 # decodes to (T, 1)
label_probs[1, 5] = 0.8 # (C, 2)
label_probs[2, 0] = 0.81 # (*, 1)
label_probs[3, 3] = 0.95 # (A, 3)
label_probs[4, 8] = 0.9 # (G, 2)
label_probs[5, 5] = 0.9 # (C, 2)
mock = common.Sample(None, None, None, None, None, label_probs)
expected = 'TCCAAAGGCC'
got = self.ls.decode_consensus(mock)
self.assertEqual(expected, got)
def haploid_sample_from_labels(ls=None,
ref=None,
pri=None,
sec=None,
pri_prob=0.6,
sec_prob=0.3):
"""Create `medaka.common.Sample` objects from a specified
reference string and predicted sequence string for easily
mocking variant calling scenarios.
"""
assert len(ref) == len(pri)
if sec is not None:
assert len(ref) == len(sec)
pos = mock_positions_array(ref)
probs = np.zeros((len(pos), len(ls._decoding)))
if sec is None:
pri_prob = pri_prob + sec_prob
for i, l in enumerate(pri):
probs[i, ls._encoding[(l, )]] = pri_prob
if sec is not None:
assert sec[i] != pri[i]
probs[i, ls._encoding[(sec[i], )]] = sec_prob
#set another label to have non-zero prob
#use the ref if is not in
#primary or secondary
other_inds = np.where(probs[i] == 0)[0]
if ls._encoding[(ref[i], )] in other_inds:
other_ind = ls._encoding[(ref[i], )]
else:
other_ind = other_inds[0]
probs[i, other_ind] = 1 - np.sum(probs[i])
s = common.Sample(ref_name='contig1',
features=None,
labels=None,
ref_seq=None,
positions=pos,
label_probs=probs)
return s, ref
def diploid_zygosity_sample_from_labels(ls=None,
ref=None,
pri=None,
sec=None,
pri_prob=None,
sec_prob=None,
het=None):
"""Create `medaka.common.Sample` objects from a specified
reference string and haplotype strings for easily
mocking variant calling scenarios.
"""
assert len(ref) == len(pri) == len(sec)
pos = mock_positions_array(ref)
probs = np.zeros((len(pos), len(ls._decoding)))
for i in range(len(ref)):
probs[i, ls._unitary_encoding[(pri[i], )]] = pri_prob
probs[i, ls._unitary_encoding[(sec[i], )]] = sec_prob
# het is always true
probs[i, -1] = int(het[i])
s = common.Sample(ref_name='contig1',
features=None,
labels=None,
ref_seq=None,
positions=pos,
label_probs=probs)
return s, ref
def test_snp_metainfo(self):
self.assertEqual(len(self.ls.snp_metainfo), 7)
def test_variant_metainfo(self):
self.assertEqual(len(self.ls.variant_metainfo), 9)