def _gen_sym_preds():
pair = {"A": "T", "T": "A", "G": "C", "C": "G"}
sym_preds = []
for f, t in "AG", "AT", "CG", "CT", "GT":
sym_preds.append(
MotifChange(f, t, forward_only=True)
| MotifChange(pair[f], pair[t], forward_only=True))
return sym_preds
def test_nr_trinuc(self):
"""This is exercising a TimeReversibleTriucleotide"""
preds = [
MotifChange("A", "C"),
MotifChange("G", "A"),
MotifChange("CGA", "TGA"),
]
sm = substitution_model.TimeReversibleTrinucleotide(predicates=preds)
got = sm.get_param_list()
self.assertEqual(got, ["A/C", "G/A", "CGA/TGA"])
self.assertEqual(len(sm.get_motifs()), 64)
def test_parse(self):
"""correctly construction"""
ag = MotifChange("A", "G")
got = parse(str(ag))
self.assertEqual(str(got), "A/G")
ts = MotifChange("A", "G") | MotifChange("C", "T")
got = parse(str(ts))
self.assertEqual(str(got), "(A/G | C/T)")
a_g = MotifChange("A", "G", forward_only=True)
t_c = MotifChange("T", "C", forward_only=True)
sym = a_g | t_c
got = parse(str(sym))
self.assertEqual(str(got), "(A>G | T>C)")
def test_getting_node_mprobs(self):
"""return correct motif probability vector for tree nodes"""
tree = make_tree(treestring="(a:.2,b:.2,(c:.1,d:.1):.1)")
aln = make_aligned_seqs(data={
"a": "TGTG",
"b": "TGTG",
"c": "TGTG",
"d": "TGTG"
})
motifs = ["T", "C", "A", "G"]
aX = MotifChange(motifs[0], motifs[3], forward_only=True).aliased("aX")
bX = MotifChange(motifs[3], motifs[0], forward_only=True).aliased("bX")
edX = MotifChange(motifs[1], motifs[2],
forward_only=True).aliased("edX")
cX = MotifChange(motifs[2], motifs[1], forward_only=True).aliased("cX")
sm = NonReversibleNucleotide(predicates=[aX, bX, edX, cX],
equal_motif_probs=True)
lf = sm.make_likelihood_function(tree)
lf.set_param_rule("aX", edge="a", value=8.0)
lf.set_param_rule("bX", edge="b", value=8.0)
lf.set_param_rule("edX", edge="edge.0", value=2.0)
lf.set_param_rule("cX", edge="c", value=0.5)
lf.set_param_rule("edX", edge="d", value=4.0)
lf.set_alignment(aln)
# we construct the hand calc variants
mprobs = ones(4, float) * 0.25
a = make_p(0.2, (0, 3), 8)
a = dot(mprobs, a)
b = make_p(0.2, (3, 0), 8)
b = dot(mprobs, b)
e = make_p(0.1, (1, 2), 2)
e = dot(mprobs, e)
c = make_p(0.1, (2, 1), 0.5)
c = dot(e, c)
d = make_p(0.1, (1, 2), 4)
d = dot(e, d)
prob_vectors = lf.get_motif_probs_by_node()
self.assertFloatEqual(prob_vectors["a"].array, a)
self.assertFloatEqual(prob_vectors["b"].array, b)
self.assertFloatEqual(prob_vectors["c"].array, c)
self.assertFloatEqual(prob_vectors["d"].array, d)
self.assertFloatEqual(prob_vectors["edge.0"].array, e)
def __init__(self, alphabet, **kw):
Stationary.__init__(self, alphabet, **kw)
alphabet = self.get_alphabet() # as may be altered by recode_gaps etc.
mask = self._instantaneous_mask
N = len(alphabet)
param_pick = numpy.zeros([N, N], int)
predicates = []
last_in_column = []
for d, (row, col) in enumerate(zip(mask, mask.T)):
row = list(numpy.flatnonzero(row[d:]) + d)
col = list(numpy.flatnonzero(col[d:]) + d)
if col:
last_in_column.append((col.pop(), d))
else:
assert not row
inst = [(d, j) for j in row] + [(i, d) for i in col]
for (i, j) in inst:
(x, y) = [alphabet[k] for k in [i, j]]
predicates.append(MotifChange(x, y, forward_only=True))
param_pick[i, j] = len(predicates)
self.param_pick = param_pick
self.last_in_column = last_in_column
predicate_masks, predicate_order = self._adapt_predicates(predicates)
self.predicate_masks = predicate_masks
self.parameter_order = []
self.predicate_indices = []
for pred in predicate_order:
mask = predicate_masks[pred]
indices = numpy.nonzero(mask)
assert numpy.alltrue(mask[indices] == 1)
self.parameter_order.append(pred)
self.predicate_indices.append(indices)
self.symmetric = False
self.check_params_exist()
def a_c(x, y):
return (x == "A" and y == "C") or (x == "C" and y == "A")
__author__ = "Peter Maxwell and Gavin Huttley"
__copyright__ = "Copyright 2007-2021, The Cogent Project"
__credits__ = ["Peter Maxwell", "Gavin Huttley"]
__license__ = "BSD-3"
__version__ = "2021.04.20a"
__maintainer__ = "Gavin Huttley"
__email__ = "*****@*****.**"
__status__ = "Production"
a_c = MotifChange("A", "C")
trans = MotifChange("A", "G") | MotifChange("T", "C")
TREE = make_tree(tip_names="ab")
class ScaleRuleTests(unittest.TestCase):
def _makeModel(self, predicates, scale_rules=None):
scale_rules = scale_rules or []
return substitution_model.TimeReversibleNucleotide(
equal_motif_probs=True,
model_gaps=False,
predicates=predicates,
scales=scale_rules,
)
Q = self.calcQ(pi, pi, *params)
P1 = FastExponentiator(Q)(0.5)
P2 = self.calc_psub_matrix(pi, 0.5, *params)
assert_allclose(P1, P2)
def _solved_nucleotide(name, predicates, rate_matrix_required=True, **kw):
if _solved_models is not None and not rate_matrix_required:
klass = PredefinedNucleotide
else:
klass = TimeReversibleNucleotide
kw["model_gaps"] = False
return klass(name=name, predicates=predicates, **kw)
kappa_y = MotifChange("T", "C").aliased("kappa_y")
kappa_r = MotifChange("A", "G").aliased("kappa_r")
kappa = (kappa_y | kappa_r).aliased("kappa")
def TN93(**kw):
"""Tamura and Nei 1993 model"""
kw["recode_gaps"] = True
return _solved_nucleotide("TN93", [kappa_y, kappa_r], **kw)
def HKY85(**kw):
"""Hasegawa, Kishino and Yanamo 1985 model"""
kw["recode_gaps"] = True
return _solved_nucleotide("HKY85", [kappa], **kw)
def _makeMotifChange(self, *args, **kw):
pred = MotifChange(*args, **kw)
return pred.interpret(self.model)
