def plot_p():
with open('../../data/split/Seq2_Sus', 'r') as f:
old_sequence = f.read()[10:].replace('\n', '').lower()
seq = models.Sequence(old_sequence)
codon_freq = codon_frequencies.F1x4(seq)
q = models.goldman_Q(codon_freq=codon_freq)
models.plot_p_over_time(q, t=0.1, codon='aaa', logscale=False)
q = models.goldman_Q()
models.plot_p_over_time(q, t=0.1, codon='aaa', logscale=False)
def plot_p():
with open('../../data/split/Seq2_Sus', 'r') as f:
old_sequence = f.read()[10:].replace('\n', '').lower()
seq = models.Sequence(old_sequence)
codon_freq = codon_frequencies.F1x4(seq)
q = models.goldman_Q(codon_freq=codon_freq)
models.plot_p_over_time(q, t=0.1, codon='aaa', logscale=False)
q = models.goldman_Q()
models.plot_p_over_time(q, t=0.1, codon='aaa', logscale=False)
def test_convert_q_to_p(self):
q = models.goldman_Q()
p = models.convert_q_to_p(q, t=0)
self.assertEqual(p.shape, (61, 61))
self.assertTrue((p==identity(61)).all())
p = models.convert_q_to_p(q, t=1)
self.assertTrue((p.max() <= 1.0))
def test_evolve_sequence_with_q(self):
q = models.goldman_Q(scale_q=True)
new_sequence0 = evolve.evolve_sequence_with_q(models.Sequence(self.old_sequence), q, t=1)
new_sequence1 = evolve.evolve_sequence_with_q(models.Sequence(self.old_sequence), q, t=1)
new_seq0 = new_sequence0.seq
new_seq1 = new_sequence1.seq
self.assertNotEqual(self.old_sequence, new_seq0) # this has a very low probability of failing
self.assertNotEqual(self.old_sequence, new_seq1) # this has a very low probability of failing
self.assertNotEqual(new_seq0, new_seq1)
def test_make_sub_from_p(self):
q = models.goldman_Q(scale_q=False)
p = models.convert_q_to_p(q, t=10)
p_cumsum, p_codons, p_cumsum_dict = models.get_cumulative_p(p, return_dict=True)
old_codon_seq = 'aaa'
old_codon = models.Codon(seq=old_codon_seq)
models.make_sub_from_p(old_codon, p_cumsum_dict)
self.assertEqual(len(old_codon.seq), 3)
allowed_letters = 'atgc'
for i in old_codon.seq:
self.assertIn(i, allowed_letters)
def test_make_subs_in_locus(self):
q = models.goldman_Q(scale_q=False)
p = models.convert_q_to_p(q, t=10)
p_cumsum, p_codons, p_cumsum_dict = models.get_cumulative_p(p, return_dict=True)
old_codon_seq = 'aaa'
old_codons = [models.Codon(seq=old_codon_seq)]*2
locus = models.Locus(codons=old_codons)
old_seq = locus.sequence
while locus.sequence == old_seq:
models.make_subs_in_locus(locus, p_cumsum_dict)
self.assertNotEqual(locus.history, [])
def evolve_tree(sequence,
taxa=10,
t=1e-2,
omega=1.0,
kappa=2.0,
lmbda=1e-5,
ti_td=0.1,
codon_freq='F1x4',
scale_q=True,
**kwargs):
"""
Evolve a parent DNA sequence into a set of daughter sequences (taxa) by:
1. generating a random phylogenetic tree
2. intantiating a mutational model (e.g. Goldman-Yang-like by default) represented by Q-matrix, with indels
3. mutate sequence according to tree shape using model
Args:
sequence: a model.Sequence instance
taxa: number of daughter sequences to evolve
t: evolution time or branch length
omega: dN/dS
kappa: ratio of transition to transversion rates
lmbda: probability of indel at codon
ti_td: ratio of insertions to deletions
codon_freq: codon frequency model, also know as equilibrium frequencies (default is F1x4)
scale_q: scales Q so that the average rate of substitution at
equilibrium equals 1. Branch lengths are thus expected number of nucleotide
substitutions per codon. See Goldman (1994).
Returns:
tree instance populated with new sequence strings
"""
codon_freq = getattr(codon_frequencies, codon_freq)(sequence)
q = models.goldman_Q(kappa=kappa,
omega=omega,
codon_freq=codon_freq,
scale_q=scale_q,
return_dict=False)
tree = trees.random_tree(taxa)
tree.value = sequence
for node in trees.get_list_of_tree_nodes(tree)[1:]:
node.value = evolve_sequence_with_q(node.parent.value,
q,
t=t,
lmbda=lmbda,
ti_td=ti_td)
return tree
def evolve_tree(sequence,
taxa=10,
t=1e-2,
omega=1.0,
kappa=2.0,
lmbda=1e-5,
ti_td=0.1,
codon_freq='F1x4',
scale_q=True,
**kwargs
):
"""
Evolve a parent DNA sequence into a set of daughter sequences (taxa) by:
1. generating a random phylogenetic tree
2. intantiating a mutational model (e.g. Goldman-Yang-like by default) represented by Q-matrix, with indels
3. mutate sequence according to tree shape using model
Args:
sequence: a model.Sequence instance
taxa: number of daughter sequences to evolve
t: evolution time or branch length
omega: dN/dS
kappa: ratio of transition to transversion rates
lmbda: probability of indel at codon
ti_td: ratio of insertions to deletions
codon_freq: codon frequency model, also know as equilibrium frequencies (default is F1x4)
scale_q: scales Q so that the average rate of substitution at
equilibrium equals 1. Branch lengths are thus expected number of nucleotide
substitutions per codon. See Goldman (1994).
Returns:
tree instance populated with new sequence strings
"""
codon_freq = getattr(codon_frequencies, codon_freq)(sequence)
q = models.goldman_Q(kappa=kappa, omega=omega, codon_freq=codon_freq, scale_q=scale_q, return_dict=False)
tree = trees.random_tree(taxa)
tree.value = sequence
for node in trees.get_list_of_tree_nodes(tree)[1:]:
node.value = evolve_sequence_with_q(node.parent.value, q, t=t, lmbda=lmbda, ti_td=ti_td)
return tree
def test_evolve_sequence_with_q(self):
q = models.goldman_Q(scale_q=True)
new_sequence0 = evolve.evolve_sequence_with_q(models.Sequence(
self.old_sequence),
q,
t=1)
new_sequence1 = evolve.evolve_sequence_with_q(models.Sequence(
self.old_sequence),
q,
t=1)
new_seq0 = new_sequence0.seq
new_seq1 = new_sequence1.seq
self.assertNotEqual(
self.old_sequence,
new_seq0) # this has a very low probability of failing
self.assertNotEqual(
self.old_sequence,
new_seq1) # this has a very low probability of failing
self.assertNotEqual(new_seq0, new_seq1)
def test_sample_model_mutation_probabilities_validation(self):
q = models.goldman_Q(scale_q=False)
sample = models.sample_model_mutation_probabilities('aaa', q)
self.assertIsInstance(sample, list)
self.assertEqual(len(sample), 100)
def test_get_mutation_from_cumulative_p(self):
q = models.goldman_Q(scale_q=False)
p = models.convert_q_to_p(q, t=10)
pc, pcod, pcdict = models.get_cumulative_p(p, return_dict=True)
old_codon = 'aaa'
new_codon = models.get_mutation_from_cumulative_p(old_codon, pcdict)
def test_get_cumulative_p(self):
q = models.goldman_Q(scale_q=False)
p = models.convert_q_to_p(q, t=10)
pc, pcod = models.get_cumulative_p(p)
for i in pc:
self.assertTrue(isclose(i[-1], 1))
def test_goldman_Q(self):
cf = codon_frequencies.FEqual()
q, qdict = models.goldman_Q(codon_freq=cf, scale_q=False, return_dict=True)
self.assertEqual(q.shape, (61, 61))
for row in q:
self.assertTrue(isclose(row.sum(), 0))
