def test_dexpmtree_nonsquareQ_assertionerror(self):
"""
Test that dexpm_tree returns an assertion error if given a non-square
Q matrix.
"""
Q = np.array([[-1, 1], [-1, 1], [1, -1]], dtype=np.double, order="C")
t = np.array([1.0])
try:
cyexpokit.dexpm_tree(Q, t)
self.fail()
except AssertionError as e:
self.assertEqual('q must be square', str(e))
def test_dexpmtree_nonsquareQ_assertionerror(self):
"""
Test that dexpm_tree returns an assertion error if given a non-square
Q matrix.
"""
Q = np.array([[-1,1], [-1,1], [1,-1]], dtype=np.double, order="C")
t = np.array([1.0])
try:
cyexpokit.dexpm_tree(Q, t)
self.fail()
except AssertionError, e:
self.assertEquals('q must be square', e.message)
def test_dexpmtree_twithzeroes_assertionerror(self):
"""
Test that dexpm_tree returns an assertion error if given a t-array
with zero or negative values
"""
Q = np.array([[-1,1,],
[1,-1,]], dtype=np.double, order="C")
t = np.array([0.0, 1.0])
try:
cyexpokit.dexpm_tree(Q, t)
self.fail()
except AssertionError, e:
self.assertEquals("All branch lengths must be greater than zero", e.message)
def test_dexpmtree_oneQtwoT2by2difT_returnsP(self):
"""
Basic test case for exponentiating a matrix and storing it in a pre-allocated
array. 2 x 2 Q matrix. List t of length 2
"""
Q = np.array([[
-1,
1,
], [
1,
-1,
]], dtype=np.double, order="C")
t = np.array([1.0, 2.0]) # Array of branch lengths
expectedP = np.array(
[
[
[0.56766765, 0.43233236], # The expected value of e^Q*1
[0.43233236, 0.56766764]
],
[
[0.50915782, 0.49084218], # The expected value of e^Q*1
[0.49084218, 0.50915782]
]
],
dtype=np.double,
order="C")
p = cyexpokit.dexpm_tree(
Q, t) # Important to note that this changes p in place
try:
np.testing.assert_allclose(expectedP, p)
except:
self.fail("expectedP != p")
def test_dexpmtree_twithzeroes_assertionerror(self):
"""
Test that dexpm_tree returns an assertion error if given a t-array
with zero or negative values
"""
Q = np.array([[
-1,
1,
], [
1,
-1,
]], dtype=np.double, order="C")
t = np.array([0.0, 1.0])
try:
cyexpokit.dexpm_tree(Q, t)
self.fail()
except AssertionError as e:
self.assertEqual("All branch lengths must be greater than zero",
str(e))
def test_hrmMk_threetiptree_matchesByHand(self):
"""
Two observed states: 0 and 1
Two hidden states per observed state: fast and slow
"""
tree = self.threetiptree
chars = [0,1,1]
# Qarray rows: 0S, 1S, 0F, 1F
# State transitions more likely than rate transitions
Q = np.array([[-.15, .1, 0.05, 0],[0.05,-.12,0,0.07],[0.06,0,-.26, .2],[0,0.08,0.3,-.38]])
t = np.array([i.length for i in tree.descendants()])
# Tips are assumed to be in both hidden states at once
# Likelihoods for tip A
L0SA = 1;L0FA = 1;L1SA = 0;L1FA = 0
# Likelhoods for tip B
L0SB = 0;L0FB = 0;L1SB = 1;L1FB = 1
# Likelihoods for tip D
L0SD = 0;L0FD = 0;L1SD = 1;L1FD = 1
p = cyexpokit.dexpm_tree(Q, t)
pvals = {}
for i,node in enumerate(["C","A","B","D"]):
for i1,state1 in enumerate(["0S","1S","0F","1F"]):
for i2,state2 in enumerate(["0S","1S","0F","1F"]):
pvals[state1 + state2 + node] = p[i,i1,i2]
L0SC = (pvals["0S0SA"] * L0SA + pvals["0S1SA"] * L1SA + pvals["0S0FA"] * L0FA + pvals["0S1FA"] * L1FA) *\
(pvals["0S0SB"] * L0SB + pvals["0S1SB"] * L1SB + pvals["0S0FB"] * L0FB + pvals["0S1FB"] * L1FB)
L0FC = (pvals["0F0SA"] * L0SA + pvals["0F1SA"] * L1SA + pvals["0F0FA"] * L0FA + pvals["0F1FA"] * L1FA) *\
(pvals["0F0SB"] * L0SB + pvals["0F1SB"] * L1SB + pvals["0F0FB"] * L0FB + pvals["0F1FB"] * L1FB)
L1SC = (pvals["1S0SA"] * L0SA + pvals["1S1SA"] * L1SA + pvals["1S0FA"] * L0FA + pvals["1S1FA"] * L1FA) *\
(pvals["1S0SB"] * L0SB + pvals["1S1SB"] * L1SB + pvals["1S0FB"] * L0FB + pvals["1S1FB"] * L1FB)
L1FC = (pvals["1F0SA"] * L0SA + pvals["1F1SA"] * L1SA + pvals["1F0FA"] * L0FA + pvals["1F1FA"] * L1FA) *\
(pvals["1F0SB"] * L0SB + pvals["1F1SB"] * L1SB + pvals["1F0FB"] * L0FB + pvals["1F1FB"] * L1FB)
L0Sr = (pvals["0S0SC"] * L0SC + pvals["0S1SC"] * L1SC + pvals["0S0FC"] * L0FC + pvals["0S1FC"] * L1FC) *\
(pvals["0S0SD"] * L0SD + pvals["0S1SD"] * L1SD + pvals["0S0FD"] * L0FD + pvals["0S1FD"] * L1FD)
L0Fr = (pvals["0F0SC"] * L0SC + pvals["0F1SC"] * L1SC + pvals["0F0FC"] * L0FC + pvals["0F1FC"] * L1FC) *\
(pvals["0F0SD"] * L0SD + pvals["0F1SD"] * L1SD + pvals["0F0FD"] * L0FD + pvals["0F1FD"] * L1FD)
L1Sr = (pvals["1S0SC"] * L0SC + pvals["1S1SC"] * L1SC + pvals["1S0FC"] * L0FC + pvals["1S1FC"] * L1FC) *\
(pvals["1S0SD"] * L0SD + pvals["1S1SD"] * L1SD + pvals["1S0FD"] * L0FD + pvals["1S1FD"] * L1FD)
L1Fr = (pvals["1F0SC"] * L0SC + pvals["1F1SC"] * L1SC + pvals["1F0FC"] * L0FC + pvals["1F1FC"] * L1FC) *\
(pvals["1F0SD"] * L0SD + pvals["1F1SD"] * L1SD + pvals["1F0FD"] * L0FD + pvals["1F1FD"] * L1FD)
predictedLikelihood = math.log(L0Sr*.25 + L0Fr*.25 + L1Sr * .25 + L1Fr *.25)
corHMMLikelihood = -2.980018
calculatedLikelihood = discrete.hrm_mk(tree, chars, Q,2, pi="Equal")
self.assertTrue(np.isclose(predictedLikelihood, calculatedLikelihood))
def test_nodelikelihood_2tiptreeSymmetricQ2x2_returnslikelihood(self):
charstates = self.charstates_01
tree = self.simpletree
for i, node in enumerate(tree.leaves()):
node.charstate = charstates[i]
Q = self.Q2x2_sym
t = [node.length for node in tree.descendants()]
t = np.array(t, dtype=np.double)
p = cyexpokit.dexpm_tree(Q, t)
for i, node in enumerate(tree.descendants()):
node.pmat = p[i]
node = tree
predictedLikelihood = 0.16483997131
calculatedLikelihood = discrete.nodeLikelihood(node)
self.assertTrue(np.isclose(predictedLikelihood, calculatedLikelihood))
def test_nodelikelihood_2tiptreeSymmetricQ2x2difblens_returnslikelihood(self):
charstates = self.charstates_01
tree = self.simpletreedifblens
for i, node in enumerate(tree.leaves()):
node.charstate = charstates[i]
Q = self.Q2x2_sym
t = [ node.length for node in tree.descendants() ]
t = np.array(t, dtype=np.double)
p = cyexpokit.dexpm_tree(Q,t)
for i, node in enumerate(tree.descendants()):
node.pmat = p[i]
node = tree
predictedLikelihood = 0.22559418195297778
calculatedLikelihood = discrete.nodeLikelihood(node)
self.assertTrue(np.isclose(predictedLikelihood, calculatedLikelihood))
def test_nodelikelihood_2tiptreesinglenodeSymmetricQ3x3_returnslikelihood(self):
charstates = self.charstates_01
tree = self.simpletree
for i, node in enumerate(tree.leaves()):
node.charstate = charstates[i]
Q = self.Q3x3_sym
t = [ node.length for node in tree.descendants() ]
t = np.array(t, dtype=np.double)
p = cyexpokit.dexpm_tree(Q,t)
for i, node in enumerate(tree.descendants()):
node.pmat = p[i]
node = tree
predictedLikelihood = 0.0863939177214389
calculatedLikelihood = discrete.nodeLikelihood(node)
self.assertTrue((predictedLikelihood, calculatedLikelihood))
def test_dexpmtree_oneQtwoT2by2difT_returnsP(self):
"""
Basic test case for exponentiating a matrix and storing it in a pre-allocated
array. 2 x 2 Q matrix. List t of length 2
"""
Q = np.array([[-1,1,],
[1,-1,]], dtype=np.double, order="C")
t = np.array([1.0,2.0]) # Array of branch lengths
expectedP = np.array([[[0.56766765, 0.43233236], # The expected value of e^Q*1
[0.43233236, 0.56766764]],
[[0.50915782, 0.49084218], # The expected value of e^Q*1
[0.49084218, 0.50915782]]],
dtype=np.double, order="C")
p = cyexpokit.dexpm_tree(Q, t) # Important to note that this changes p in place
try:
np.testing.assert_allclose(expectedP, p)
except:
self.fail("expectedP != p")
def test_mkFlatroot_2tiptreeSymmetricQ2x2difblens_returnslikelihood(self):
charstates = self.charstates_01
tree = self.simpletreedifblens
for i, node in enumerate(tree.leaves()):
node.charstate = charstates[i]
Q = self.Q2x2_sym
t = [node.length for node in tree.descendants()]
t = np.array(t, dtype=np.double)
p = cyexpokit.dexpm_tree(Q, t)
for i, node in enumerate(tree.descendants()):
node.pmat = p[i]
node = tree
predictedLikelihood = math.log(0.11279709097648889)
calculatedLikelihood = discrete.mk(tree, charstates, Q)
self.assertTrue(np.isclose(predictedLikelihood, calculatedLikelihood))
def test_mkFlatroot_2tiptreeSymmetricQ2x2difblens_returnslikelihood(self):
charstates = self.charstates_01
tree = self.simpletreedifblens
for i, node in enumerate(tree.leaves()):
node.charstate = charstates[i]
Q = self.Q2x2_sym
t = [ node.length for node in tree.descendants() ]
t = np.array(t, dtype=np.double)
p = cyexpokit.dexpm_tree(Q,t)
for i, node in enumerate(tree.descendants()):
node.pmat = p[i]
node = tree
predictedLikelihood = math.log(0.11279709097648889)
calculatedLikelihood = discrete.mk(tree, charstates, Q)
self.assertTrue(np.isclose(predictedLikelihood, calculatedLikelihood))
import unittest
import ivy
from ivy.chars.expokit import cyexpokit
import numpy as np
import random
Q = np.array([[-1,1,0,0],
[0,-1,1,0],
[0,0,-1,1],
[0,0,0,0]], dtype=np.double, order='C')
t = np.array([ random.random() for x in range(10000) ])
%timeit -n2 -r3 cyexpokit.dexpm_tree(Q, t)
def test_hrmMk_threetiptree_matchesByHand(self):
"""
Two observed states: 0 and 1
Two hidden states per observed state: fast and slow
"""
tree = self.threetiptree
chars = [0, 1, 1]
# Qarray rows: 0S, 1S, 0F, 1F
# State transitions more likely than rate transitions
Q = np.array([[-.15, .1, 0.05, 0], [0.05, -.12, 0, 0.07],
[0.06, 0, -.26, .2], [0, 0.08, 0.3, -.38]])
t = np.array([i.length for i in tree.descendants()])
# Tips are assumed to be in both hidden states at once
# Likelihoods for tip A
L0SA = 1
L0FA = 1
L1SA = 0
L1FA = 0
# Likelhoods for tip B
L0SB = 0
L0FB = 0
L1SB = 1
L1FB = 1
# Likelihoods for tip D
L0SD = 0
L0FD = 0
L1SD = 1
L1FD = 1
p = cyexpokit.dexpm_tree(Q, t)
pvals = {}
for i, node in enumerate(["C", "A", "B", "D"]):
for i1, state1 in enumerate(["0S", "1S", "0F", "1F"]):
for i2, state2 in enumerate(["0S", "1S", "0F", "1F"]):
pvals[state1 + state2 + node] = p[i, i1, i2]
L0SC = (pvals["0S0SA"] * L0SA + pvals["0S1SA"] * L1SA + pvals["0S0FA"] * L0FA + pvals["0S1FA"] * L1FA) *\
(pvals["0S0SB"] * L0SB + pvals["0S1SB"] * L1SB + pvals["0S0FB"] * L0FB + pvals["0S1FB"] * L1FB)
L0FC = (pvals["0F0SA"] * L0SA + pvals["0F1SA"] * L1SA + pvals["0F0FA"] * L0FA + pvals["0F1FA"] * L1FA) *\
(pvals["0F0SB"] * L0SB + pvals["0F1SB"] * L1SB + pvals["0F0FB"] * L0FB + pvals["0F1FB"] * L1FB)
L1SC = (pvals["1S0SA"] * L0SA + pvals["1S1SA"] * L1SA + pvals["1S0FA"] * L0FA + pvals["1S1FA"] * L1FA) *\
(pvals["1S0SB"] * L0SB + pvals["1S1SB"] * L1SB + pvals["1S0FB"] * L0FB + pvals["1S1FB"] * L1FB)
L1FC = (pvals["1F0SA"] * L0SA + pvals["1F1SA"] * L1SA + pvals["1F0FA"] * L0FA + pvals["1F1FA"] * L1FA) *\
(pvals["1F0SB"] * L0SB + pvals["1F1SB"] * L1SB + pvals["1F0FB"] * L0FB + pvals["1F1FB"] * L1FB)
L0Sr = (pvals["0S0SC"] * L0SC + pvals["0S1SC"] * L1SC + pvals["0S0FC"] * L0FC + pvals["0S1FC"] * L1FC) *\
(pvals["0S0SD"] * L0SD + pvals["0S1SD"] * L1SD + pvals["0S0FD"] * L0FD + pvals["0S1FD"] * L1FD)
L0Fr = (pvals["0F0SC"] * L0SC + pvals["0F1SC"] * L1SC + pvals["0F0FC"] * L0FC + pvals["0F1FC"] * L1FC) *\
(pvals["0F0SD"] * L0SD + pvals["0F1SD"] * L1SD + pvals["0F0FD"] * L0FD + pvals["0F1FD"] * L1FD)
L1Sr = (pvals["1S0SC"] * L0SC + pvals["1S1SC"] * L1SC + pvals["1S0FC"] * L0FC + pvals["1S1FC"] * L1FC) *\
(pvals["1S0SD"] * L0SD + pvals["1S1SD"] * L1SD + pvals["1S0FD"] * L0FD + pvals["1S1FD"] * L1FD)
L1Fr = (pvals["1F0SC"] * L0SC + pvals["1F1SC"] * L1SC + pvals["1F0FC"] * L0FC + pvals["1F1FC"] * L1FC) *\
(pvals["1F0SD"] * L0SD + pvals["1F1SD"] * L1SD + pvals["1F0FD"] * L0FD + pvals["1F1FD"] * L1FD)
predictedLikelihood = math.log(L0Sr * .25 + L0Fr * .25 + L1Sr * .25 +
L1Fr * .25)
corHMMLikelihood = -2.980018
calculatedLikelihood = discrete.hrm_mk(tree, chars, Q, 2, pi="Equal")
self.assertTrue(np.isclose(predictedLikelihood, calculatedLikelihood))
