def test_mutate(self):
"""Probs mutate should return correct vector from input vector"""
a = Alphabet('abc')**2
m = Probs([0.5, 0.25, 0.25, 0.1, 0.8, 0.1, 0.3, 0.6, 0.1], a)
#because of fp math in accumulate, can't predict boundaries exactly
#so add/subtract eps to get the result we expect
eps = 1e-6
# a b b a c c a b c
seq = array([0, 1, 1, 0, 2, 2, 0, 1, 2])
random_vec = array([0, .01, .8 - eps, 1, 1, .3, .05, .9 + eps, .95])
self.assertEqual(m.mutate(seq, random_vec), \
# a a b c c a a c c
array([0,0,1,2,2,0,0,2,2]))
#check that freq. distribution is about right
seqs = array([m.mutate(seq) for i in range(1000)])
#WARNING: bool operators return byte arrays, whose sums wrap at 256!
zero_count = asarray(seqs == 0, 'int32')
sums = sum(zero_count, axis=0)
#expect: 500, 100, 100, 500, 300, 300, 500, 100, 300
#std dev = sqrt(npq), which is sqrt(250), sqrt(90), sqrt(210)
means = array([500, 100, 100, 500, 300, 300, 500, 100, 300])
var = array([250, 90, 90, 250, 210, 210, 250, 90, 210])
three_sd = 3 * sqrt(var)
for obs, exp, sd in zip(sums, means, three_sd):
assert exp - 2 * sd < obs < exp + 2 * sd
def test_fixNegsReflect(self):
"""Rates fixNegsReflect should reflect negatives across diagonal"""
ab = Alphabet('ab')**2
#should leave matrix alone if no off-diagonal elements
q = Rates([0, 0, 1, -1], ab)
self.assertEqual(q.fixNegsReflect()._data, array([[0, 0], [1, -1]]))
q = Rates([-2, 2, 1, -1], ab)
self.assertEqual(q.fixNegsReflect()._data, array([[-2, 2], [1, -1]]))
#should work if precisely one off-diag element in a pair is negative
q = Rates([2, -2, 1, -1], ab)
self.assertEqual(q.fixNegsReflect()._data, array([[0, 0], [3, -3]]))
q = Rates([-1, 1, -2, 2], ab)
self.assertEqual(q.fixNegsReflect()._data, array([[-3, 3], [0, -0]]))
#should work if both off-diag elements in a pair are negative
q = Rates([1, -1, -2, 2], ab)
self.assertEqual(q.fixNegsReflect()._data, array([[-2, 2], [1, -1]]))
q = Rates([2, -2, -1, 1], ab)
self.assertEqual(q.fixNegsReflect()._data, array([[-1, 1], [2, -2]]))
q = Rates(
[[0, 3, -2, -1], [2, -1, 2, -3], [-1, -1, 2, 0], [-3, 2, 0, 1]],
RnaPairs)
q2 = q.fixNegsReflect()
self.assertEqual(q2._data, \
array([[-7, 3, 1, 3],
[ 2, -5, 3, 0],
[ 2, 0, -2, 0],
[ 1, 5, 0, -6]]))
def test_toCounts(self):
"""Probs toCounts should return counts object w/ right numbers"""
a = Alphabet('abc')**2
m = Probs([0.5, 0.25, 0.25, 0.1, 0.8, 0.1, 0.3, 0.6, 0.1], a)
obs = m.toCounts(30)
assert isinstance(obs, Counts)
exp = Counts([[5., 2.5, 2.5, 1, 8, 1, 3, 6, 1]], a)
self.assertEqual(obs, exp)
def test_makeModel(self):
"""Probs makeModel should return correct substitution pattern"""
a = Alphabet('abc')**2
m = Probs([0.5, 0.25, 0.25, 0.1, 0.8, 0.1, 0.3, 0.6, 0.1], a)
obs = m.makeModel(array([0, 1, 1, 0, 2, 2]))
exp = array([[0.5,0.25,0.25],[0.1,0.8,0.1],[0.1,0.8,0.1],\
[0.5,0.25,0.25],[0.3,0.6,0.1],[0.3,0.6,0.1]])
self.assertEqual(obs, exp)
def setUp(self):
"""Defines some standard test items."""
self.ab = Alphabet('ab')
class abUsage(Usage):
Alphabet = self.ab
self.abUsage = abUsage
def test_getitem(self):
"""PairMatrix __getitem__ should translate indices and get from array"""
n = self.named
self.assertEqual(n['a'], array([1, 2]))
self.assertEqual(n['b'], array([3, 4]))
self.assertEqual(n['a', 'a'], 1)
self.assertEqual(n['a', 'b'], 2)
self.assertEqual(n['b', 'a'], 3)
self.assertEqual(n['b', 'b'], 4)
#WARNING: m[a][b] doesn't work b/c indices not translated!
#must access as m[a,b] instead.
filterwarnings("ignore", "using a non-integer")
try:
x = n['a']['b']
except (ValueError, IndexError):
pass
#should work even if SubAlphabets not the same
a = Alphabet('ab')
x = Alphabet('xyz')
j = a * x
m = PairMatrix([1, 2, 3, 4, 5, 6], j)
self.assertEqual(m['a', 'x'], 1)
self.assertEqual(m['a', 'y'], 2)
self.assertEqual(m['a', 'z'], 3)
self.assertEqual(m['b', 'x'], 4)
self.assertEqual(m['b', 'y'], 5)
self.assertEqual(m['b', 'z'], 6)
#should work even if SubAlphabets are different types
a = Alphabet([1, 2, 3])
b = Alphabet(['abc', 'xyz'])
j = a * b
m = PairMatrix([1, 2, 3, 4, 5, 6], j)
self.assertEqual(m[1, 'abc'], 1)
self.assertEqual(m[1, 'xyz'], 2)
self.assertEqual(m[2, 'abc'], 3)
self.assertEqual(m[2, 'xyz'], 4)
self.assertEqual(m[3, 'abc'], 5)
self.assertEqual(m[3, 'xyz'], 6)
self.assertEqual(list(m[2]), [3, 4])
#gives KeyError if single item not present in first level
self.assertRaises(KeyError, m.__getitem__, 'x')
def test_toRates(self):
"""Probs toRates should return log of probs, optionally normalized"""
a = Alphabet('abc')**2
p = Probs([0.9, 0.05, 0.05, 0.1, 0.85, 0.05, 0.02, 0.02, 0.96], a)
assert p.isValid()
r = p.toRates()
assert isinstance(r, Rates)
assert r.isValid()
assert not r.isComplex()
self.assertEqual(r._data, logm(p._data))
r_norm = p.toRates(normalize=True)
self.assertFloatEqual(trace(r_norm._data), -1.0)
def test_ne(self):
"""PairMatrix test for inequality should check all elements"""
p = self.ab_pairs
a = PairMatrix.empty(p)
b = PairMatrix.empty(p)
c = PairMatrix([1, 2, 3, 4], p)
d = PairMatrix([1, 2, 3, 4], p)
assert a != c
assert a == b
assert c == d
#Note: still compare equal if alphabets are different
x = Alphabet('xy')
x = x * x
y = PairMatrix([1, 2, 3, 4], x)
assert y == c
#should check all elements, not just first
c = PairMatrix([1, 1, 1, 1], p)
d = PairMatrix([1, 1, 1, 4], p)
assert c != d
def test_eq(self):
"""Pairmatrix test for equality should check all elements"""
p = self.ab_pairs
a = PairMatrix.empty(p)
b = PairMatrix.empty(p)
assert a is not b
self.assertEqual(a, b)
c = PairMatrix([1, 2, 3, 4], p)
d = PairMatrix([1, 2, 3, 4], p)
assert c is not d
self.assertEqual(c, d)
self.assertNotEqual(a, c)
#Note: still compare equal if alphabets are different
x = Alphabet('xy')
x = x * x
y = PairMatrix([1, 2, 3, 4], x)
self.assertEqual(y, c)
#should check all elements, not just first
c = PairMatrix([1, 1, 1, 1], p)
d = PairMatrix([1, 1, 1, 4], p)
assert c is not d
self.assertNotEqual(c, d)
def _with(self, motifs):
a = Alphabet._with(self, motifs)
a.__class__ = type(self)
a._gc = self._gc
return a
def __init__(self, motifset, Gap=IUPAC_gap, Missing=IUPAC_missing,\
Gaps=None,
Sequence=None, Ambiguities=None,
label=None, Complements=None, Pairs=None, MWCalculator=None, \
add_lower=False, preserve_existing_moltypes=False, \
make_alphabet_group=False, ModelSeq=None):
"""Returns a new MolType object. Note that the parameters are in flux.
Currently:
motifset: Alphabet or sequence of items in the default
alphabet. Does not include degenerates.
Gap: default gap symbol
Missing: symbol for missing data
Gaps: any other symbols that should be treated as gaps (doesn't have
to include Gap or Missing; they will be silently added)
Sequence: Class for constructing sequences.
Ambiguities: dict of char:tuple, doesn't include gaps (these are
hard-coded as - and ?, and added later.
label: text label, don't know what this is used for. Unnecessary?
Complements: dict of symbol:symbol showing how the non-degenerate
single characters complement each other. Used for constructing
on the fly the complement table, incl. support for mustPair and
canPair.
Pairs: dict in which keys are pairs of symbols that can pair
with each other, values are True (must pair) or False (might
pair). Currently, the meaning of GU pairs as 'weak' is conflated
with the meaning of degenerate symbol pairs (which might pair
with each other but don't necessarily, depending on how the
symbol is resolved). This should be refactored.
MWCalculator: f(seq) -> molecular weight.
add_lower: if True (default: False) adds the lowercase versions of
everything into the alphabet. Slated for deletion.
preserve_existing_moltypes: if True (default: False), does not
set the MolType of the things added in **kwargs to self.
make_alphabet_group: if True, makes an AlphabetGroup relating
the various alphabets to one another.
ModelSeq: sequence type for modeling
Note on "Degenerates" versus "Ambiguities": self.Degenerates contains
_only_ mappings for degenerate symbols, whereas self.Ambiguities
contains mappings for both degenerate and non-degenerate symbols.
Sometimes you want one, sometimes the other, so both are provided.
"""
self.Gap = Gap
self.Missing = Missing
self.Gaps = frozenset([Gap, Missing])
if Gaps:
self.Gaps = self.Gaps.union(frozenset(Gaps))
self.label = label
#set the sequence constructor
if Sequence is None:
Sequence = ''.join #safe default string constructor
elif not preserve_existing_moltypes:
Sequence.MolType = self
self.Sequence = Sequence
#set the ambiguities
ambigs = {self.Missing:tuple(motifset)+(self.Gap,),self.Gap:(self.Gap,)}
if Ambiguities:
ambigs.update(Ambiguities)
for c in motifset:
ambigs[c] = (c,)
self.Ambiguities = ambigs
#set Complements -- must set before we make the alphabet group
self.Complements = Complements or {}
if make_alphabet_group: #note: must use _original_ ambiguities here
self.Alphabets = AlphabetGroup(motifset, Ambiguities, \
MolType=self)
self.Alphabet = self.Alphabets.Base
else:
if isinstance(motifset, Enumeration):
self.Alphabet = motifset
elif max(len(motif) for motif in motifset) == 1:
self.Alphabet = CharAlphabet(motifset, MolType=self)
else:
self.Alphabet = Alphabet(motifset, MolType=self)
#set the other properties
self.Degenerates = Ambiguities and Ambiguities.copy() or {}
self.Degenerates[self.Missing] = ''.join(motifset)+self.Gap
self.Matches = make_matches(motifset, self.Gaps, self.Degenerates)
self.Pairs = Pairs and Pairs.copy() or {}
self.Pairs.update(make_pairs(Pairs, motifset, self.Gaps, \
self.Degenerates))
self.MWCalculator = MWCalculator
#add lowercase characters, if we're doing that
if add_lower:
self._add_lowercase()
#cache various other data that make the calculations faster
self._make_all()
self._make_comp_table()
# a gap can be a true gap char or a degenerate character, typically '?'
# we therefore want to ensure consistent treatment across the definition
# of characters as either gap or degenerate
self.GapString = ''.join(self.Gaps)
strict_gap = "".join(set(self.GapString) - set(self.Degenerates))
self.stripDegenerate = FunctionWrapper(
keep_chars(strict_gap+''.join(self.Alphabet)))
self.stripBad = FunctionWrapper(keep_chars(''.join(self.All)))
to_keep = set(self.Alphabet) ^ set(self.Degenerates) - set(self.Gaps)
self.stripBadAndGaps = FunctionWrapper(keep_chars(''.join(to_keep)))
#make inverse degenerates from degenerates
#ensure that lowercase versions also exist if appropriate
inv_degens = {}
for key, val in self.Degenerates.items():
inv_degens[frozenset(val)] = key.upper()
if add_lower:
inv_degens[frozenset(''.join(val).lower())] = key.lower()
for m in self.Alphabet:
inv_degens[frozenset(m)] = m
if add_lower:
inv_degens[frozenset(''.join(m).lower())] = m.lower()
for m in self.Gaps:
inv_degens[frozenset(m)] = m
self.InverseDegenerates = inv_degens
#set array type for modeling alphabets
try:
self.ArrayType = self.Alphabet.ArrayType
except AttributeError:
self.ArrayType = None
#set modeling sequence
self.ModelSeq = ModelSeq
def _with(self, motifs):
a = Alphabet._with(self, motifs)
a.__class__ = type(self)
a._gc = self._gc
return a
def setUp(self):
"""Define standard alphabets."""
self.abc = Alphabet('abc')
self.abc_pairs = self.abc**2
def setUp(self):
"""Define an alphabet and some probs."""
self.ab = Alphabet('ab')
self.ab_pairs = self.ab**2
def setUp(self):
"""Define standard alphabet and matrices for tests."""
self.ab = Alphabet('ab')
self.ab_pairs = self.ab * self.ab
self.empty = PairMatrix([0, 0, 0, 0], self.ab_pairs)
self.named = PairMatrix([[1, 2], [3, 4]], self.ab_pairs, 'name')
from cogent.maths.matrix_logarithm import logm
from cogent.maths.matrix_exponentiation import FastExponentiator as expm
#need to find test directory to get access to the tests of the Freqs interface
try:
from os import getcwd
from sys import path
from os.path import sep, join
test_path = getcwd().split(sep)
index = test_path.index('tests')
fields = test_path[:index + 1] + ["test_maths"]
test_path = sep + join(*fields)
path.append(test_path)
from test_stats.test_util import StaticFreqsTestsI
my_alpha = Alphabet('abcde')
class myUsage(Usage):
Alphabet = my_alpha
class UsageAsFreqsTests(StaticFreqsTestsI, TestCase):
"""Note that the remaining Usage methods are tested here."""
ClassToTest = myUsage
except ValueError: #couldn't find directory
pass
__author__ = "Rob Knight"
__copyright__ = "Copyright 2007-2016, The Cogent Project"
__credits__ = ["Rob Knight", "Daniel McDonald"]
__license__ = "GPL"
