def test_EstimateDistances_fromUnaligned(self):
"""Excercising estimate distances from unaligned sequences"""
d = EstimateDistances(self.collection, JC69(), do_pair_align=True,
rigorous_align=True)
d.run()
canned_result = {('b', 'e'): 0.440840,
('c', 'e'): 0.440840,
('a', 'c'): 0.088337,
('a', 'b'): 0.188486,
('a', 'e'): 0.440840,
('b', 'c'): 0.0883373}
result = d.getPairwiseDistances()
self.assertDistsAlmostEqual(canned_result, result)
d = EstimateDistances(self.collection, JC69(), do_pair_align=True,
rigorous_align=False)
d.run()
canned_result = {('b', 'e'): 0.440840,
('c', 'e'): 0.440840,
('a', 'c'): 0.088337,
('a', 'b'): 0.188486,
('a', 'e'): 0.440840,
('b', 'c'): 0.0883373}
result = d.getPairwiseDistances()
self.assertDistsAlmostEqual(canned_result, result)
def get_alignment(tree, N_sites):
'build a PyCogent alignment object from a tree and length'
sm = JC69()
lf = sm.makeLikelihoodFunction(tree)
lf.setConstantLengths()
aln = lf.simulateAlignment(sequence_length=N_sites)
return (aln)
def ml4(aln, true_tree):
'''
Input a true tree and an alignment
Calculate the likelihood of all possible unrooted 4-taxon trees
Return True if the ML tree is the true tree
Return False otherwise
'''
# all trees with unit branch lengths
all_trees = [
LoadTree(treestring='((a,b),(c,d))'),
LoadTree(treestring='((a,c),(b,d))'),
LoadTree(treestring='((a,d),(b,c))')
]
# optimise lf for all trees
sm = JC69()
results = []
for t in all_trees:
lf = sm.makeLikelihoodFunction(t)
lf.setAlignment(aln)
lf.optimise(local=True)
results.append(lf.getLogLikelihood())
# get the ml tree and compare to true tree
ml_tree = all_trees[results.index(max(results))]
return ml_tree.sameTopology(true_tree)
def test_EstimateDistances_fromThreeway(self):
"""testing (well, exercising at least), EsimateDistances fromThreeway"""
d = EstimateDistances(self.al, JC69(), threeway=True)
d.run()
canned_result = {('b', 'e'): 0.495312,
('c', 'e'): 0.479380,
('a', 'c'): 0.089934,
('a', 'b'): 0.190021,
('a', 'e'): 0.495305,
('b', 'c'): 0.0899339}
result = d.getPairwiseDistances(summary_function="mean")
self.assertDistsAlmostEqual(canned_result, result)
def evaluate_tree(aln):
d = distance.EstimateDistances(aln, submodel=JC69())
d.run(show_progress=False)
njtree = nj.nj(d.getPairwiseDistances())
if debug:
print(d)
print(njtree.asciiArt())
print(njtree.sameTopology(tr))
for otu in 'BCD':
print(njtree.getConnectingEdges('A', otu))
L = njtree.getConnectingEdges('A', 'B')
return len(L) == 3
def optimization(result, aln, tree1, tree2):
# get the sites for each tree according to the assignments
aln1 = LoadSeqs(data=[('a', ''), ('c', ''), ('b', ''), ('d', '')],
moltype=DNA)
aln2 = LoadSeqs(data=[('a', ''), ('c', ''), ('b', ''), ('d', '')],
moltype=DNA)
for i in range(len(aln)):
if (result[i] == 1):
aln1 = aln1 + aln[i]
if (result[i] == 2):
aln2 = aln2 + aln[i]
if (result[i] == 0):
aln1 = aln1 + aln[i]
aln2 = aln2 + aln[i]
tree_parameter = [[], []]
modle = JC69()
# calculate the likelihood and do optimization. optimise will generates
# new tree parameters
lf1 = modle.makeLikelihoodFunction(tree1)
lf1.setAlignment(aln1)
lf1.optimise(local=True)
likelihood1 = lf1.getLogLikelihood()
# new tree parameters generates by optimise. As tree1/2 is symmetric, get
# p,q,r from 6 branch lengths
p1 = (lf1.getParamValue('length', 'a') +
lf1.getParamValue('length', 'c')) / 2.0
q1 = (lf1.getParamValue('length', 'b') +
lf1.getParamValue('length', 'd')) / 2.0
r1 = lf1.getParamValue('length', 'edge.1') + \
lf1.getParamValue('length', 'edge.0')
lf2 = modle.makeLikelihoodFunction(tree2)
lf2.setAlignment(aln2)
lf2.optimise(local=True)
likelihood2 = lf2.getLogLikelihood()
p2 = (lf2.getParamValue('length', 'a') +
lf2.getParamValue('length', 'c')) / 2.0
q2 = (lf2.getParamValue('length', 'b') +
lf2.getParamValue('length', 'd')) / 2.0
r2 = lf2.getParamValue('length', 'edge.1') + \
lf2.getParamValue('length', 'edge.0')
# return the new tree_parameter. As likelihood is in log, so plus together
# get the total likelihood for the whole sites
tree_parameter[0] = [p1, q1, r1]
tree_parameter[1] = [p2, q2, r2]
likelihood = likelihood1 + likelihood2
return tree_parameter, likelihood
def expectation_singlesite(aln, tree1, tree2):
modle = JC69()
result = []
lf1 = modle.makeLikelihoodFunction(tree1)
lf2 = modle.makeLikelihoodFunction(tree2)
for i in range(len(aln)):
lf1.setAlignment(aln[i])
prob1 = lf1.getLogLikelihood()
lf2.setAlignment(aln[i])
prob2 = lf2.getLogLikelihood()
if (prob1 > prob2):
result.append(1)
if (prob1 < prob2):
result.append(2)
if (prob1 == prob2):
result.append(0)
return result
def test_EstimateDistances(self):
"""testing (well, exercising at least), EstimateDistances"""
d = EstimateDistances(self.al, JC69())
d.run()
canned_result = {('b', 'e'): 0.440840,
('c', 'e'): 0.440840,
('a', 'c'): 0.088337,
('a', 'b'): 0.188486,
('a', 'e'): 0.440840,
('b', 'c'): 0.0883373}
result = d.getPairwiseDistances()
self.assertDistsAlmostEqual(canned_result, result)
# excercise writing to file
d.writeToFile('junk.txt')
try:
os.remove('junk.txt')
except OSError:
pass # probably parallel
def optimization(result, aln, tree1, tree2, tree):
aln1 = LoadSeqs(data=[('a', ''), ('c', ''), ('b', ''), ('d', '')],
moltype=DNA)
aln2 = LoadSeqs(data=[('a', ''), ('c', ''), ('b', ''), ('d', '')],
moltype=DNA)
for i in range(len(aln)):
if (result[i] == 1):
aln1 = aln1 + aln[i]
if (result[i] == 2):
aln2 = aln2 + aln[i]
if (result[i] == 0):
aln1 = aln1 + aln[i]
aln2 = aln2 + aln[i]
print "aln1:"
print aln1
print "aln2:"
print aln2
tree_parameter = [[], []]
modle = JC69()
lf1 = modle.makeLikelihoodFunction(tree)
lf1.setAlignment(aln1)
lf1.optimise(local=True)
p1 = (lf1.getParamValue('length', 'a') +
lf1.getParamValue('length', 'c')) / 2.0
q1 = (lf1.getParamValue('length', 'b') +
lf1.getParamValue('length', 'd')) / 2.0
r1 = lf1.getParamValue('length', 'edge.1') + lf1.getParamValue(
'length', 'edge.0')
lf2 = modle.makeLikelihoodFunction(tree)
lf2.setAlignment(aln2)
lf2.optimise(local=True)
p2 = (lf2.getParamValue('length', 'a') +
lf2.getParamValue('length', 'c')) / 2.0
q2 = (lf2.getParamValue('length', 'b') +
lf2.getParamValue('length', 'd')) / 2.0
r2 = lf2.getParamValue('length', 'edge.1') + lf2.getParamValue(
'length', 'edge.0')
tree_parameter[0] = [p1, q1, r1]
tree_parameter[1] = [p2, q2, r2]
print "tree parameter in this time"
print tree_parameter
return tree_parameter
def expectation_singlesite(aln, tree1, tree2):
modle = JC69()
result = []
lf1 = modle.makeLikelihoodFunction(tree1)
# make lilelihood function for tree1,tree2
lf2 = modle.makeLikelihoodFunction(tree2)
for i in range(
len(aln)
): # for each site,compare the likelihood for it belong to tree1,tree2
# and assign it to the one with larger likelihood
lf1.setAlignment(aln[i])
prob1 = lf1.getLogLikelihood()
lf2.setAlignment(aln[i])
prob2 = lf2.getLogLikelihood()
# if(prob1 > prob2):
# result.append(1)
# if(prob1 < prob2):
# result.append(2)
# # if it is the same, assign to both of the two trees
# if(prob1 == prob2):
# result.append(0)
_max = max(prob1, prob2)
prob1 -= _max
prob2 -= _max
exp_prob1 = np.exp(prob1)
exp_prob2 = np.exp(prob2)
_sum = exp_prob1 + exp_prob2
random_prob = np.random.random() * _sum
if exp_prob1 < random_prob:
result.append(1)
else:
result.append(2)
return result