def test_multi_alpha_diagonalness_of_blockdiagonal_blocks(self):
from multicov.alignment import Alignment
from multicov.alphabet import protein_alphabet, dna_alphabet, rna_alphabet
from multicov.align_io import load_fasta
from multicov.binary import binary_index_map
from multicov.statistics import Statistics, MaxentModel
from os.path import join
align1 = load_fasta(join('test_data', 'test_aln1.fasta'),
protein_alphabet,
invalid_letter_policy='gap')
align2 = load_fasta(join('test_data', 'test_aln2.fasta'),
dna_alphabet,
invalid_letter_policy='gap')
align3 = load_fasta(join('test_data', 'test_aln2.fasta'),
rna_alphabet,
invalid_letter_policy='gap')
align = Alignment(align1)
align.add(align2).add(align3)
stats = Statistics(align, regularization_amount=0.5)
maxent = MaxentModel(stats)
bin_map = binary_index_map(stats)
for crt_range in bin_map:
crt_slice = slice(crt_range[0], crt_range[1])
crt_block = maxent.couplings[crt_slice, crt_slice]
self.assertLess(
np.max(np.abs(crt_block - np.diag(np.diag(crt_block)))), 1e-10)
def test_multi_alpha_shape_and_symmetry_of_couplings(self):
from multicov.alignment import Alignment
from multicov.alphabet import protein_alphabet, dna_alphabet, rna_alphabet
from multicov.align_io import load_fasta
from multicov.statistics import Statistics, MaxentModel
from os.path import join
align1 = load_fasta(join('test_data', 'test_aln2.fasta'),
dna_alphabet,
invalid_letter_policy='gap')
align2 = load_fasta(join('test_data', 'test_aln1.fasta'),
protein_alphabet,
invalid_letter_policy='gap')
align3 = load_fasta(join('test_data', 'test_aln2.fasta'),
rna_alphabet,
invalid_letter_policy='uppergap')
align = Alignment(align1)
align.add(align2).add(align3)
stats = Statistics(align, regularization_amount=0.5)
maxent = MaxentModel(stats)
self.assertLess(np.max(np.abs(maxent.couplings - maxent.couplings.T)),
1e-10)
self.assertSequenceEqual(
np.shape(maxent.couplings), 2 * [
4 * (align1.get_width() + align3.get_width()) +
20 * align2.get_width()
])
def test_freq2_on_multi_alpha(self):
from multicov.alphabet import protein_alphabet, dna_alphabet
from multicov.align_io import load_fasta
from multicov.statistics import Statistics
from os.path import join
align = load_fasta(join('test_data', 'test_aln2.fasta'),
dna_alphabet,
invalid_letter_policy='gap')
align2 = load_fasta(join('test_data', 'test_aln1.fasta'),
protein_alphabet,
invalid_letter_policy='gap')
align.add(align2)
stats = Statistics(align)
expected_f2 = _slow_get_freq2(align)
self.assertTrue(np.allclose(stats.freq2, expected_f2))
def test_mask_upper(self):
from multicov.alignment import ReferenceMapping
from multicov.align_io import load_fasta
from multicov.alphabet import protein_alphabet
align = load_fasta(os.path.join('test_data', 'test_aln3.fasta'),
protein_alphabet,
invalid_letter_policy='upper',
mask_fct='upper')
align0 = load_fasta(os.path.join('test_data', 'test_aln3.fasta'),
protein_alphabet,
invalid_letter_policy='unchanged')
mask = [not _.islower() and _ != '.' for _ in align0[0, :]]
expected = align0.truncate_columns(mask)
expected.reference = ReferenceMapping(
list(range(expected.data.shape[1])))
self.assertEqual(align, expected)
def test_replace_invalid_by_uppercase_then_leave(self):
from multicov.alignment import Alignment
from multicov.align_io import load_fasta
from multicov.alphabet import dna_alphabet
align = load_fasta(os.path.join('test_data', 'test_aln2.fasta'),
dna_alphabet,
invalid_letter_policy='upper')
expected = Alignment(['GATTACA', 'ACCA--T', 'G.C-A-C'], dna_alphabet)
expected.annotations['name'] = ['one', 'sequence', 'one line']
self.assertEqual(align, expected)
def test_dna_unchanged_invalid(self):
from multicov.alignment import Alignment
from multicov.align_io import load_fasta
from multicov.alphabet import dna_alphabet
align = load_fasta(os.path.join('test_data', 'test_aln2.fasta'),
dna_alphabet,
invalid_letter_policy='unchanged')
expected = Alignment(['GATTACA', 'ACCA--T', 'G.c-a-c'], dna_alphabet)
expected.annotations['name'] = ['one', 'sequence', 'one line']
self.assertEqual(align, expected)
def test_multi_alpha(self):
from multicov.alphabet import protein_alphabet, dna_alphabet
from multicov.align_io import load_fasta
from multicov.statistics import Statistics
from os.path import join
align = load_fasta(join('test_data', 'test_aln2.fasta'),
dna_alphabet,
invalid_letter_policy='gap')
align2 = load_fasta(join('test_data', 'test_aln1.fasta'),
protein_alphabet,
invalid_letter_policy='gap')
align.add(align2)
alpha = 0.6
stats1 = Statistics(align)
stats2 = Statistics(align,
regularization_amount=alpha,
regularizer='pseudocount')
bkg_freq1 = np.hstack(
np.ones(width * alphabet.size(no_gap=True)) / alphabet.size()
for alphabet, width in align.alphabets)
bkg_freq2 = np.outer(bkg_freq1, bkg_freq1)
freq1 = (1 - alpha) * stats1.freq1 + alpha * bkg_freq1
freq2 = (1 - alpha) * stats1.freq2 + alpha * bkg_freq2
n_letts = np.hstack(width * [alphabet.size(no_gap=True)]
for alphabet, width in align.alphabets)
idxs0 = np.hstack(([0], np.cumsum(n_letts)[:-1]))
for idx0, n_lett in zip(idxs0, n_letts):
idxs = slice(idx0, idx0 + n_lett)
# noinspection PyUnresolvedReferences
freq2[idxs, idxs] = np.diag(freq1[idxs])
cmat = freq2 - np.outer(freq1, freq1)
self.assertTrue(np.allclose(freq1, stats2.freq1))
# noinspection PyTypeChecker
self.assertTrue(np.allclose(freq2, stats2.freq2))
# noinspection PyTypeChecker
self.assertTrue(np.allclose(cmat, stats2.cmat))
def test_protein_unchanged_invalid(self):
from multicov.alignment import Alignment
from multicov.align_io import load_fasta
from multicov.alphabet import protein_alphabet
align = load_fasta(os.path.join('test_data', 'test_aln1.fasta'),
protein_alphabet,
invalid_letter_policy='unchanged')
expected = Alignment(['IVGGYTCQ', 'XVGGTEAQ', 'IGG-KDT-'],
alphabet=protein_alphabet)
expected.annotations['name'] = ['seq1', 'seq2', 'seq3']
self.assertEqual(align, expected)
def test_protein_keep_annot_ws(self):
from multicov.alignment import Alignment
from multicov.align_io import load_fasta
from multicov.alphabet import protein_alphabet
align = load_fasta(os.path.join('test_data', 'test_aln1.fasta'),
protein_alphabet,
strip_ws_in_annot=False)
expected = Alignment(['IVGGYTCQ', '-VGGTEAQ', 'IGG-KDT-'],
alphabet=protein_alphabet)
expected.annotations['name'] = ['seq1 ', ' seq2', 'seq3']
self.assertEqual(align, expected)
def test_cmat_on_multi_alpha(self):
from multicov.alphabet import protein_alphabet, dna_alphabet, rna_alphabet
from multicov.align_io import load_fasta
from multicov.statistics import Statistics
from os.path import join
align = load_fasta(join('test_data', 'test_aln2.fasta'),
dna_alphabet,
invalid_letter_policy='gap')
align2 = load_fasta(join('test_data', 'test_aln1.fasta'),
protein_alphabet,
invalid_letter_policy='gap')
align3 = load_fasta(join('test_data', 'test_aln2.fasta'),
rna_alphabet,
invalid_letter_policy='uppergap')
align.add(align2).add(align3)
stats = Statistics(align)
expected_f1 = _slow_get_freq1(align)
expected_f2 = _slow_get_freq2(align)
expected_cmat = expected_f2 - np.outer(expected_f1, expected_f1)
self.assertTrue(np.allclose(stats.cmat, expected_cmat))
def test_gap_gauge(self):
from multicov.alignment import Alignment
from multicov.alphabet import protein_alphabet, dna_alphabet, rna_alphabet
from multicov.align_io import load_fasta
from multicov.statistics import Statistics, MaxentModel
from os.path import join
align1 = load_fasta(join('test_data', 'test_aln2.fasta'),
dna_alphabet,
invalid_letter_policy='gap')
align2 = load_fasta(join('test_data', 'test_aln1.fasta'),
protein_alphabet,
invalid_letter_policy='gap')
align3 = load_fasta(join('test_data', 'test_aln2.fasta'),
rna_alphabet,
invalid_letter_policy='uppergap')
align = Alignment(align1)
align.add(align2).add(align3)
stats = Statistics(align, regularization_amount=0.5)
maxent = MaxentModel(stats)
energies = maxent.score([align.get_width() * '-'])
self.assertLess(np.max(np.abs(energies)), 1e-10)
def test_mask_from_first_seq(self):
from multicov.alignment import Alignment
from multicov.align_io import load_fasta
from multicov.alphabet import protein_alphabet
from numpy import in1d
align = load_fasta(os.path.join('test_data', 'test_aln1.fasta'),
protein_alphabet,
invalid_letter_policy='unchanged',
mask_fct=lambda s: ~in1d(list(s), ['V', 'G']))
expected = Alignment(['IYTCQ', 'XTEAQ', 'IKDT-'],
alphabet=protein_alphabet)
expected.annotations['name'] = ['seq1', 'seq2', 'seq3']
self.assertEqual(align, expected)