def test_against_matlab_example(self):
from multicov.alignment import Alignment
from multicov.alphabet import protein_alphabet
from multicov.statistics import Statistics, MaxentModel
from scipy.io import loadmat
import os.path
matlab = loadmat(os.path.join('test_data', 'maxent_sample.mat'),
squeeze_me=True)
align = Alignment(matlab['alignment']['data'][()], protein_alphabet)
stats = Statistics(align, regularization_amount=0.5)
maxent = MaxentModel(stats)
energies = maxent.score(align)
self.assertTrue(np.allclose(energies, matlab['energies'][()]))
def test_with_matrix(self):
from multicov.alignment import Alignment
from multicov.binary import binary_index_map
from multicov.alphabet import protein_alphabet
from multicov.statistics import Statistics, MaxentModel
align = Alignment([
'WKHNAY', 'KHRCDA', 'LGVVGY', 'LIGDDH', 'CMPRYW', 'QWFWRA',
'VTMPEG', 'LNYINM', 'WHV-EW', 'PIWGGF', 'PPCWVE', 'E-MWRG',
'RFGKFT', 'CGRCGS', 'T-PMVW', 'LNCPYA'
], protein_alphabet)
stats = Statistics(align, regularization_amount=0.1)
maxent = MaxentModel(stats)
energies = maxent.score(align.data)
energies0 = maxent.score(align)
self.assertTrue(np.allclose(energies, energies0))
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_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_on_empty(self):
from multicov.alignment import Alignment
from multicov.statistics import Statistics, MaxentModel
stats = Statistics(Alignment())
maxent = MaxentModel(stats)
self.assertTrue(hasattr(maxent, 'stats'))
self.assertTrue(hasattr(maxent, 'alphabets'))
self.assertTrue(hasattr(maxent, 'annotations'))
self.assertTrue(hasattr(maxent, 'reference'))
self.assertIs(maxent.stats, stats)
self.assertIs(maxent.alphabets, stats.alphabets)
self.assertIs(maxent.annotations, stats.annotations)
self.assertIs(maxent.reference, stats.reference)
self.assertEqual(np.size(maxent.couplings), 0)
def test_against_matlab_example(self):
from multicov.alignment import Alignment
from multicov.alphabet import protein_alphabet
from multicov.statistics import Statistics, MaxentModel
from scipy.io import loadmat
import os.path
matlab = loadmat(os.path.join('test_data', 'maxent_sample.mat'),
squeeze_me=True)
align = Alignment(matlab['alignment']['data'][()], protein_alphabet)
stats = Statistics(align, regularization_amount=0.5)
maxent = MaxentModel(stats)
self.assertTrue(np.allclose(stats.cmat, matlab['dca']['cmat'][()]))
self.assertTrue(
np.allclose(
maxent.couplings,
matlab['params_nogap_nofc_nodiagtrick']['couplings'][()]))
def test_on_protein(self):
from multicov.alignment import Alignment
from multicov.binary import binary_index_map
from multicov.alphabet import protein_alphabet
from multicov.statistics import Statistics, MaxentModel
align = Alignment([
'WKHNAY', 'KHRCDA', 'LGVVGY', 'LIGDDH', 'CMPRYW', 'QWFWRA',
'VTMPEG', 'LNYINM', 'WHV-EW', 'PIWGGF', 'PPCWVE', 'E-MWRG',
'RFGKFT', 'CGRCGS', 'T-PMVW', 'LNCPYA'
], protein_alphabet)
stats = Statistics(align, regularization_amount=0.1)
maxent = MaxentModel(stats)
invC = -np.linalg.inv(stats.cmat)
bin_map = binary_index_map(stats)
for crt_range in bin_map:
invC[crt_range[0]:crt_range[1], crt_range[0]:crt_range[1]] = 0
self.assertTrue(
np.allclose(invC,
maxent.couplings - np.diag(np.diag(maxent.couplings))))