def relative_entropy_terms(self, background=None):
"""
Computes a row-wise relative entropy terms per motif and stores them in a DictArray.
Parameters
----------
background : dict
{motif_1: prob_1, motif_2: prob_2, ...} is the specified background distribution.
Returns
-------
DictArray
Notes
-----
If background is type None, it defaults to equifrequent.
"""
if background is None:
num_motifs = len(self.motifs)
background = array([1 / num_motifs] * num_motifs)
else:
background = array([background.get(m, 0) for m in self.motifs])
validate_freqs_array(background)
ret = background * (safe_log(background) - safe_log(self.array))
return self.template.wrap(ret)
def __init__(self, data, motifs, row_indices=None):
super(MotifFreqsArray, self).__init__(data,
motifs,
row_indices,
dtype=float)
axis = 0 if self.array.ndim == 1 else 1
validate_freqs_array(self.array, axis=axis)
def __init__(self, data, motifs, row_indices=None, background=None):
data = numpy.array(data)
row_sum = data.sum(axis=1)
# are we dealing with counts data?
if 0 <= data.min() and 1 < data.max():
# convert to freqs data
data = data / numpy.vstack(row_sum)
row_sum = data.sum(axis=1)
# are we dealing with freqs data?
if (data >= 0).all() and numpy.allclose(
row_sum[numpy.isnan(row_sum) == False], 1
):
# standard PSSM object creation
if background is None:
background = numpy.ones(len(motifs), dtype=float) / len(motifs)
self._background = numpy.array(background)
assert len(background) == len(
motifs
), "Mismatch between number of motifs and the background"
validate_freqs_array(self._background)
pssm = safe_log(data) - safe_log(self._background)
super(PSSM, self).__init__(
pssm, motifs, row_indices=row_indices, dtype=float
)
self._indices = numpy.arange(self.shape[0]) # used for scoring
return
if not (data.min() < 0 < data.max()):
raise ValueError("PSSM has been supplied invalid data")
# we dealing with pssm data
super(PSSM, self).__init__(data, motifs, row_indices=row_indices, dtype=float)
self._indices = numpy.arange(self.shape[0]) # used for scoring
def jsd(freqs1, freqs2, validate=False):
"""calculate Jensen–Shannon divergence between two probability distributions
Parameters
----------
freqs1 : one dimensional array
row vector frequencies, sum to 1
freqs2 : one dimensional array
row vector frequencies, sum to 1
validate : bool
"""
# Convert input arrays into numpy arrays
freqs1 = array(freqs1)
freqs2 = array(freqs2)
if validate:
assert_equal(freqs1.shape,
freqs2.shape,
err_msg="freqs1/freqs2 mismatched shape")
assert freqs1.ndim == 1, "freqs1 has incorrect dimension"
assert freqs2.ndim == 1, "freqs2 has incorrect dimension"
try:
validate_freqs_array(freqs1)
validate_freqs_array(freqs2)
except ValueError as err:
raise AssertionError("freqs not valid") from err
H_mn = safe_p_log_p(freqs1 / 2 + freqs2 / 2).sum()
mn_H = sum([sum(i) for i in map(safe_p_log_p, [freqs1, freqs2])]) / 2
return H_mn - mn_H
def jsd(freqs1, freqs2, validate=False):
"""calculate Jensen–Shannon divergence between two probability distributions
Parameters
----------
freqs1 : one dimensional array
row vector frequencies, sum to 1
freqs2 : one dimensional array
row vector frequencies, sum to 1
validate : bool
"""
# Convert input arrays into numpy arrays
freqs1 = array(freqs1)
freqs2 = array(freqs2)
if validate:
assert_equal(freqs1.shape,
freqs2.shape,
err_msg="freqs1/freqs2 mismatched shape")
assert freqs1.ndim == 1, "freqs1 has incorrect dimension"
assert freqs2.ndim == 1, "freqs2 has incorrect dimension"
try:
validate_freqs_array(freqs1)
validate_freqs_array(freqs2)
except ValueError as err:
raise AssertionError("freqs not valid") from err
H_mn = fsum(safe_p_log_p(freqs1 / 2 + freqs2 / 2))
mn_H = fsum([fsum(i) for i in map(safe_p_log_p, [freqs1, freqs2])]) / 2
jsd_ = H_mn - mn_H
if jsd_ < 0 and isclose(jsd_, 0, atol=1e-10):
jsd_ = 0
elif jsd_ < 0:
raise ArithmeticError(
f"{jsd_} is negative and below defined precision threshold")
return jsd_
