def from_counts(cls, alphabet, counts, prior= None):
"""Build a LogoData object from counts."""
# Counts is a Motif object?
#counts = counts.array
seq_length, A = counts.shape
if prior is not None: prior = array(prior, float64)
if prior is None or sum(prior)==0.0:
R = log(A)
ent = zeros( seq_length, float64)
entropy_interval = None
for i in range (0, seq_length) :
C = sum(counts[i])
#FIXME: fixup corebio.moremath.entropy()?
if C == 0 :
ent[i] = 0.0
else :
ent[i] = R - entropy(counts[i])
else :
ent = zeros( seq_length, float64)
entropy_interval = zeros( (seq_length,2) , float64)
R = log(A)
for i in range (0, seq_length) :
alpha = array(counts[i] , float64)
alpha += prior
posterior = Dirichlet(alpha)
ent[i] = posterior.mean_relative_entropy(prior/sum(prior))
entropy_interval[i][0], entropy_interval[i][1] = \
posterior.interval_relative_entropy(prior/sum(prior), 0.95)
weight = array( na.sum(counts,axis=1) , float)
weight /= max(weight)
return cls(seq_length, alphabet, counts, ent, entropy_interval, weight)
def from_counts(cls, alphabet, counts, prior= None):
"""Build a LogoData object from counts."""
# Counts is a Motif object?
#counts = counts.array
seq_length, A = counts.shape
if prior is not None: prior = array(prior, float64)
if prior is None or sum(prior)==0.0:
R = log(A)
ent = zeros( seq_length, float64)
entropy_interval = None
for i in range (0, seq_length) :
C = sum(counts[i])
#FIXME: fixup corebio.moremath.entropy()?
if C == 0 :
ent[i] = 0.0
else :
ent[i] = R - entropy(counts[i])
else :
ent = zeros( seq_length, float64)
entropy_interval = zeros( (seq_length,2) , float64)
R = log(A)
for i in range (0, seq_length) :
alpha = array(counts[i] , float64)
alpha += prior
posterior = Dirichlet(alpha)
ent[i] = posterior.mean_relative_entropy(prior/sum(prior))
entropy_interval[i][0], entropy_interval[i][1] = \
posterior.interval_relative_entropy(prior/sum(prior), 0.95)
weight = array( na.sum(counts,axis=1) , float)
weight /= max(weight)
return cls(seq_length, alphabet, counts, ent, entropy_interval, weight)
def from_counts(cls, alphabet, counts, stats_func=None, prior=None, composition=None,
ngdata=None, pvalue=None):
"""Build a LogoData object from counts."""
# Counts is a Motif object?
#counts = counts.array
if alphabet in [codon_dna_alphabet, codon_rna_alphabet]:
seq_length, A = len(counts), len(alphabet)
else:
seq_length, A = counts.shape
if prior is not None: prior = array(prior, float64)
if ngdata is not None:
if ngdata.counts.shape[0] != seq_length:
raise ValueError("Sequence length in negative dataset should be the same as the input's.")
if prior is None or sum(prior)==0.0:
R = log(A)
odds = None
ent = zeros(seq_length, float64)
entropy_interval = None
max_value = 0.0
for i in range (0, seq_length) :
C = sum(counts[i])
if C == 0:
ent[i] = 0.0
else:
ent[i] = R - entropy(counts[i])
if max_value < na.max(ent[i]):
max_value = na.max(ent[i])
else :
odds = []
ent = zeros(seq_length, float64)
entropy_interval = zeros( (seq_length,2) , float64)
max_value_ent = 0.0
max_value_int = 0.0
refdata = None
if pvalue is not None or stats_func is None:
pvalue = array(pvalue, float64)
pvalue_calc = False
else:
pvalue = []
pvalue_calc = True
R = log(A)
for i in range (0, seq_length):
alpha = array(counts[i] , float64)
alpha += prior
posterior = Dirichlet(alpha)
ent[i] = posterior.mean_relative_entropy(prior/sum(prior))
odds.append(posterior.odds_ratio(composition))
entropy_interval[i][0], entropy_interval[i][1] = \
posterior.interval_relative_entropy(prior/sum(prior), 0.95)
if pvalue_calc == True:
if ngdata is not None:
ngalpha = array(ngdata.counts[i], float64)
ngalpha += ngdata.prior
ngposterior = Dirichlet(ngalpha)
ngcompos = ngposterior.alpha/sum(ngposterior.alpha)
pvalue.append( stats_func(posterior.alpha, ngcompos) )
else:
pvalue.append( stats_func(posterior.alpha, composition) )
if max_value_ent < na.max(ent[i]):
max_value_ent = na.max(ent[i])
if max_value_int < na.max(entropy_interval[i][1]):
max_value_int = na.max(entropy_interval[i][1])
max_value = max_value_ent + max_value_int
weight = array( na.sum(counts,axis=1) , float)
weight /= max(weight)
return cls(seq_length, alphabet, counts, ent, entropy_interval,
weight, pvalue, composition, odds, max_value, prior)
