def signal_diff(self, scale, feature_ids=False):
# TODO scale
num_scales = 19
if not(feature_ids):
feat_vec = numpy.zeros(num_scales)
for index in xrange(num_scales):
scale = sequtil.get_georgiev_scales()[index]
feat_vec[index] = self.mutation_signal_distance(scale)
return feat_vec
else:
ids = ['%i' % (i) for i in xrange(num_scales)]
names = ['Georgiev %i signal difference' % (i)
for i in xrange(num_scales)]
return (ids, names)
def _parse_scales(self, scales):
if (type(scales) == str):
try:
scales = int(scales)
except ValueError:
pass
# retrieve the set of Georgiev aa scales
if (scales == 'gg'):
scale_list = sequtil.get_georgiev_scales()
scale_ids = ['gg%i' % (i) for i in xrange(1, len(scale_list) + 1)]
scale_names = [
'Georgiev scale %i' % (i)
for i in xrange(1,
len(scale_list) + 1)
]
# retrieve AAIndex scale with index scales
elif (type(scales) == int):
scale_list = [sequtil.get_aaindex_scale(scales)]
scale_ids = ['aai%i' % (scales)]
scale_names = ['amino acid index %i' % (scales)]
# retrieve list of pseaac scale
elif (scales[0] == 'p' and len(scales) > 2):
scale_indices = [int(i) for i in scales.split('p')[1:]]
scale_list = scale_indices
scale_ids = ['pseaac%i' % (i + 1) for i in scale_indices]
scale_names = ['PseAAC scale %i' % (i + 1) for i in scale_indices]
# retrieve pseaac scale
elif (scales[0] == 'p'):
scale_index = int(scales[1:])
scale_list = [int(scales[1:])]
scale_ids = ['pseaac%i' % (scale_index + 1)]
scale_names = ['PseAAC scale %i' % (scale_index + 1)]
# retrieve list of AAIndex scales... (still used somewhere?)
elif (type(scales) == list and all([type[i] == int for i in scales])):
scale_list = [sequtil.get_aaindex_scale(i) for i in scales]
scale_ids = ['aai%i' % (i) for i in scales]
scale_names = ['amino acid index %i' % (i) for i in scales]
else:
raise ValueError('Incorrect scale provided: %s\n' % (str(scales)))
return (scale_list, scale_ids, scale_names)
def msa_signal_diff(self, scale, feature_ids=False):
# TODO scale
num_scales = 19
if not(feature_ids):
feat_vec = numpy.zeros(num_scales)
for index in xrange(num_scales):
scale = sequtil.get_georgiev_scales()[index]
feat_vec[index] = self.min_signal_dist_to_msa(scale)
return feat_vec
else:
ids = ['%i' % (i) for i in xrange(num_scales)]
names = ['Georgiev %i signal dist. to msa variability' % (i)
for i in xrange(num_scales)]
return (ids, names)
def _parse_scales(self, scales):
if(type(scales) == str):
try:
scales = int(scales)
except ValueError:
pass
# retrieve the set of Georgiev aa scales
if(scales == 'gg'):
scale_list = sequtil.get_georgiev_scales()
scale_ids = ['gg%i' % (i) for i in xrange(1, len(scale_list) + 1)]
scale_names = ['Georgiev scale %i' % (i)
for i in xrange(1, len(scale_list) + 1)]
# retrieve list of pseaac scale
elif(scales[0] == 'p' and len(scales) > 2):
scale_indices = [int(i) for i in scales.split('p')[1:]]
scale_list = scale_indices
scale_ids = ['pseaac%i' % (i + 1) for i in scale_indices]
scale_names = ['PseAAC scale %i' % (i + 1) for i in scale_indices]
# retrieve pseaac scale
elif(scales[0] == 'p'):
scale_index = int(scales[1:])
scale_list = [int(scales[1:])]
scale_ids = ['pseaac%i' % (scale_index + 1)]
scale_names = ['PseAAC scale %i' % (scale_index + 1)]
# retrieve AAIndex scale with index scales
elif(type(scales) == int):
scale_list = [sequtil.get_aaindex_scale(scales)]
scale_ids = ['aai%i' % (scales)]
scale_names = ['amino acid index %i' % (scales)]
# retrieve list of AAIndex scales... (still used somewhere?)
elif(type(scales) == list and all([type[i] == int for i in scales])):
scale_list = [sequtil.get_aaindex_scale(i) for i in scales]
scale_ids = ['aai%i' % (i) for i in scales]
scale_names = ['amino acid index %i' % (i) for i in scales]
else:
raise ValueError('Incorrect scale provided: %s\n' % (str(scales)))
return (scale_list, scale_ids, scale_names)
def signal_auc(self, scale, env_window=21, sig_window=9, edge=1.0,
threshold=1.5, below_threshold=False,
feature_ids=False):
# TODO scale
num_scales = 19
if not(feature_ids):
feat_vec = numpy.zeros(num_scales)
for index in xrange(num_scales):
scale = sequtil.get_georgiev_scales()[index]
auc = self.environment_signal_peak_area(
env_window, scale, sig_window, edge, threshold,
below_threshold)
# anscombe transform (~poissos --> ~normal)
feat_vec[index] = 2 * numpy.sqrt(auc + (3.0 / 8.0))
return feat_vec
else:
ids = ['%i' % (i) for i in xrange(num_scales)]
names = ['Georgiev %i signal ew%i sw%i e%.2f th%.2f' %
(i, env_window, sig_window, edge, threshold)
for i in xrange(num_scales)]
return (ids, names)
