def test_interval_to_array(self):
"""Testing interval to array code"""
ref = "chr1"
endpos = 1000
chunkstep = 10
dtype = "i"
# test different interval sizes
for intervalsize in xrange(1,100,10):
fullarr = np.zeros((endpos+intervalsize,1), dtype=dtype)
intervals = []
val = 0
for start in xrange(0, endpos, intervalsize):
intervals.append(Interval(ref, start, start+intervalsize, POS_STRAND, val))
fullarr[start:start+intervalsize,0] = val
val += 2
# test different chunk sizes
for chunksize in xrange(chunkstep, endpos, chunkstep):
testarr = np.zeros((endpos+intervalsize,1), dtype=dtype)
write_interval_data_to_array(iter(intervals),
{"chr1": testarr},
dtype=dtype,
chunksize=chunksize,
num_channels=1,
channel_dict=get_channel_dict())
self.assertTrue(np.all(testarr == fullarr))
#
# test intervals on different chromosomes
#
refiter = itertools.cycle(itertools.chain(itertools.repeat("chr1", 3),
itertools.repeat("chr2", 3) ,
itertools.repeat("chr3", 3)))
for intervalsize in (1,10,50,100):
#print "intervalsize", intervalsize
fullarr = {"chr1":np.zeros((endpos+intervalsize,1), dtype=dtype),
"chr2":np.zeros((endpos+intervalsize,1), dtype=dtype),
"chr3":np.zeros((endpos+intervalsize,1), dtype=dtype)}
intervals = []
val = 0
for start in xrange(0, endpos, intervalsize):
ref = refiter.next()
intervals.append(Interval(ref, start, start+intervalsize, POS_STRAND, val))
fullarr[ref][start:start+intervalsize] = val
val += 2
# test different chunk sizes
for chunksize in (1, 8, 16, 32, 64, 128, 256, 512, 1024):
testarr = {"chr1":np.zeros((endpos+intervalsize,1), dtype=dtype),
"chr2":np.zeros((endpos+intervalsize,1), dtype=dtype),
"chr3":np.zeros((endpos+intervalsize,1), dtype=dtype)}
write_interval_data_to_array(iter(intervals),
testarr,
dtype=dtype,
chunksize=chunksize,
num_channels=1,
channel_dict=get_channel_dict())
for chrom in testarr:
self.assertTrue(np.all(testarr[chrom] == fullarr[chrom]))
def test_get_channel_dict(self):
# default
d = get_channel_dict(is_pe=False, is_strand=False, is_allele=False)
for v in d.values():
self.assertEqual(v, (0,))
# paired end
d = get_channel_dict(is_pe=True, is_strand=False, is_allele=False)
self.assertEqual(d[(0,NO_STRAND,"A")], (0,))
self.assertEqual(d[(0,POS_STRAND,"G")], (0,))
self.assertEqual(d[(1,NEG_STRAND,"C")], (1,))
self.assertEqual(d[(1,POS_STRAND,"T")], (1,))
self.assertEqual(d[(None,NEG_STRAND,"N")], (0,1))
# strand
d = get_channel_dict(is_pe=False, is_strand=True, is_allele=False)
self.assertEqual(d[(0,NO_STRAND,"A")], (0,1))
self.assertEqual(d[(1,POS_STRAND,"G")], (0,))
self.assertEqual(d[(0,NEG_STRAND,"C")], (1,))
# allele
d = get_channel_dict(is_pe=False, is_strand=False, is_allele=True)
self.assertEqual(d[(0,NO_STRAND,"A")], (0,))
self.assertEqual(d[(1,POS_STRAND,"G")], (1,))
self.assertEqual(d[(0,NEG_STRAND,"C")], (2,))
self.assertEqual(d[(0,NEG_STRAND,"T")], (3,))
self.assertEqual(d[(0,NO_STRAND,"N")], (0,1,2,3,))
# pe/strand
d = get_channel_dict(is_pe=True, is_strand=True, is_allele=False)
self.assertEqual(d[(None,NO_STRAND,"A")], (0,1,2,3))
self.assertEqual(d[(1,POS_STRAND,"G")], (1,))
self.assertEqual(d[(0,NEG_STRAND,"C")], (2,))
self.assertEqual(d[(1,NEG_STRAND,"T")], (3,))
self.assertEqual(d[(None,NEG_STRAND,"N")], (2,3))
# pe/allele
d = get_channel_dict(is_pe=True, is_strand=False, is_allele=True)
self.assertEqual(d[(0,NO_STRAND,"A")], (0,))
self.assertEqual(d[(0,POS_STRAND,"G")], (2,))
self.assertEqual(d[(1,NEG_STRAND,"C")], (5,))
self.assertEqual(d[(1,POS_STRAND,"T")], (7,))
self.assertEqual(d[(None,NEG_STRAND,"N")], (0,1,2,3,4,5,6,7))
# strand/allele
d = get_channel_dict(is_pe=False, is_strand=True, is_allele=True)
self.assertEqual(d[(None,NO_STRAND,"A")], (0,1))
self.assertEqual(d[(1,POS_STRAND,"G")], (2,))
self.assertEqual(d[(0,NEG_STRAND,"C")], (5,))
self.assertEqual(d[(1,NEG_STRAND,"T")], (7,))
self.assertEqual(d[(None,NEG_STRAND,"N")], (1,3,5,7))
# pe/strand/allele
d = get_channel_dict(is_pe=True, is_strand=True, is_allele=True)
self.assertEqual(d[(None,NO_STRAND,"A")], (0,1,2,3))
self.assertEqual(d[(1,POS_STRAND,"G")], (5,))
self.assertEqual(d[(0,NEG_STRAND,"C")], (10,))
self.assertEqual(d[(1,NO_STRAND,"T")], (13,15))
self.assertEqual(d[(None,NO_STRAND,"N")], tuple(range(0,16)))
self.assertEqual(d[(None,NEG_STRAND,"N")], (2,3,6,7,10,11,14,15))
self.assertEqual(d[(1,POS_STRAND,"N")], (1,5,9,13))
def test_stranded_allele_intervals(self):
"""testing coverage with allele frequencies"""
dtype = "f"
channel_dict = get_channel_dict(False, True, True)
pos_strand_channels = channel_dict[(None,POS_STRAND,None)]
neg_strand_channels = channel_dict[(None,NEG_STRAND,None)]
intervals1, correct1 = \
random_stranded_allele_intervals(100, self.length,
self.isize_max, dtype)
total_cov = correct1.sum()
t = self.tf.create_track("a", VectorTrack, strand=True, allele=True)
# test loading from intervals
t.fromintervals(iter(intervals1))
self.assertTrue(np.all(t["gene1"] == correct1))
# test count function
intervals2, correct2 = random_intervals(10, self.length, self.isize_max, dtype)
for ival in intervals2:
ref = ival.ref
start = ival.start
end = ival.end
strand = ival.strand
val = ival.value
# check plus strand
mycount = t.count((ref, start, end, POS_STRAND, val))
correctcount = correct1[start:end,pos_strand_channels].sum()
self.assertAlmostEqual(mycount, correctcount)
mycov = t.coverage((ref, start, end, POS_STRAND, val), multiplier=1.0)
correctcov = correct1[start:end,pos_strand_channels].sum(axis=1) / float(total_cov)
self.assertTrue(np.allclose(mycov, correctcov, atol=1e-4))
mydens = t.density((ref, start, end, POS_STRAND, val), multiplier=1.0)
correctdens = correctcount / float(total_cov * (end - start))
self.assertTrue(np.allclose(mydens, correctdens, atol=1e-4))
# check minus strand
mycount = t.count((ref, start, end, NEG_STRAND, val))
correctcount = correct1[start:end,neg_strand_channels].sum()
self.assertAlmostEqual(mycount, correctcount)
mycov = t.coverage((ref, start, end, NEG_STRAND, val), multiplier=1.0)
correctcov = correct1[start:end,neg_strand_channels].sum(axis=1) / float(total_cov)
self.assertTrue(np.allclose(mycov, correctcov, atol=1e-4))
mydens = t.density((ref, start, end, NEG_STRAND, val), multiplier=1.0)
correctdens = correctcount / float(total_cov * (end - start))
self.assertTrue(np.allclose(mydens, correctdens, atol=1e-4))
# check both strands
mycount = t.count((ref, start, end, NO_STRAND, val))
correctcount = correct1[start:end].sum()
self.assertAlmostEqual(mycount, correctcount)
mycov = t.coverage((ref, start, end, NO_STRAND, val), multiplier=1.0)
correctcov = correct1[start:end].sum(axis=1) / float(total_cov)
self.assertTrue(np.allclose(mycov, correctcov, atol=1e-4))
mydens = t.density((ref, start, end, NO_STRAND, val), multiplier=1.0)
correctdens = correctcount / float(total_cov * (end - start))
self.assertTrue(np.allclose(mydens, correctdens, atol=1e-4))
def random_stranded_allele_intervals(n, length, isize_max, dtype):
channel_dict = get_channel_dict(False, True, True)
intervals = []
correct = np.zeros((length,8), dtype)
for i in xrange(n):
start = np.random.randint(0, length-isize_max)
end = start + np.random.randint(1, isize_max)
seq = []
for x in xrange(start, end):
dna = random.choice("ATGCN")
pos_channels = channel_dict[(None,POS_STRAND,dna)]
neg_channels = channel_dict[(None,NEG_STRAND,dna)]
correct[x,pos_channels] += 2.0 / len(pos_channels)
correct[x,neg_channels] += -1.0 / len(neg_channels)
seq.append(dna)
seq = ''.join(seq)
intervals.append(SequenceInterval('gene1', start, end, POS_STRAND, 2, seq=seq))
intervals.append(SequenceInterval('gene1', start, end, NEG_STRAND, -1, seq=seq))
return intervals, correct
