def load_LatticeProteins(local_env, path=''):
all_data = Proteins_utils.load_FASTA(path + 'Lattice_Proteins_MSA.fasta')
B = all_data.shape[0]
seed = utilities.check_random_state(0)
shuffle = np.argsort(seed.rand(B))
train_data = all_data[shuffle][:int(0.8 * B)]
test_data = all_data[shuffle][int(0.8 * B):]
train_weights = None
test_weights = None
true_contacts_A = [(21, 12), (26, 3), (9, 6), (18, 1), (27, 8), (25, 18),
(27, 16), (20, 1), (19, 12), (13, 10), (15, 8),
(19, 14), (7, 4), (17, 14), (25, 16), (26, 7), (23, 2),
(23, 20), (24, 19), (24, 15), (15, 10), (24, 7),
(24, 11), (23, 4), (22, 11), (22, 5), (25, 2), (11, 6)]
contact_map = np.zeros([27, 27])
for i, j in true_contacts_A:
contact_map[i - 1, j - 1] = 1
contact_map[j - 1, i - 1] = 1
local_env['train_data'] = train_data.astype(curr_int)
local_env['test_data'] = test_data.astype(curr_int)
local_env['train_weights'] = train_weights
local_env['test_weights'] = test_weights
local_env['contact_map'] = contact_map.astype(curr_int)
def load_WW(local_env, path=''):
all_data = Proteins_utils.load_FASTA(path + 'WW_domain_MSA.fasta')
num_neighbours = Proteins_utils.count_neighbours(all_data)
all_weights = 1.0 / num_neighbours
B = all_data.shape[0]
seed = utilities.check_random_state(0)
shuffle = np.argsort(seed.rand(B))
train_data = all_data[shuffle][:int(0.8 * B)]
test_data = all_data[shuffle][int(0.8 * B):]
train_weights = all_weights[shuffle][:int(0.8 * B)]
test_weights = all_weights[shuffle][int(0.8 * B):]
env = pickle.load(open(path + 'WW_test_sequences.data', 'rb'),
encoding=encoding)
experimental_data = np.asarray(np.concatenate([
env['sequences_1'], env['sequences_2'], env['sequences_3'],
env['sequences_4']
],
axis=0),
dtype='int')
experimental_labels = np.asarray(np.concatenate([
np.zeros(len(env['sequences_1'])),
np.ones(len(env['sequences_2']) + len(env['sequences_3'])),
2 * np.ones(len(env['sequences_4']))
],
axis=0),
dtype='int')
local_env['train_data'] = train_data.astype(curr_int)
local_env['test_data'] = test_data.astype(curr_int)
local_env['train_weights'] = train_weights.astype(curr_float)
local_env['test_weights'] = test_weights.astype(curr_float)
local_env['experimental_data'] = experimental_data.astype(curr_int)
local_env['experimental_labels'] = experimental_labels.astype(curr_int)
def __init__(self,
N=100,
nature='Bernoulli',
n_c=1,
random_state=None,
gauge='zerosum',
zero_field=False):
self.N = N
self.nature = nature
self.random_state = utilities.check_random_state(random_state)
if self.nature == 'Potts':
self.n_c = n_c
else:
self.n_c = 1
self.zero_field = zero_field
self.interpolate = False
super(BM, self).__init__(n_layers=1,
layers_size=[self.N],
layers_nature=[self.nature + '_coupled'],
layers_n_c=[self.n_c],
layers_name=['layer'])
self.gauge = gauge
self.layer = layer.initLayer(N=self.N,
nature=self.nature + '_coupled',
position='visible',
n_c=self.n_c,
random_state=self.random_state,
zero_field=self.zero_field,
gauge=self.gauge)
self.init_couplings(0.01)
self.tmp_l2_fields = 0
def load_Audition_souris(local_env, path=''):
from scipy.io import loadmat
all_data = np.asarray(loadmat(path + 'Audition.mat')['binNinf_double_seg'],
dtype='int')
B = all_data.shape[0]
seed = utilities.check_random_state(0)
shuffle = np.argsort(seed.rand(B))
train_data = all_data[shuffle][:int(0.8 * B)]
test_data = all_data[shuffle][int(0.8 * B):]
train_weights = None
test_weights = None
local_env['train_data'] = train_data.astype(curr_int)
local_env['test_data'] = test_data.astype(curr_int)
local_env['train_weights'] = train_weights
local_env['test_weights'] = test_weights
def load_Hsp70(local_env, path=''):
all_data, all_labels = Proteins_utils.load_FASTA(path +
'Hsp70_protein_MSA.fasta',
with_labels=True)
all_weights = pickle.load(open(path + 'Hsp70_info.data', 'rb'),
encoding=encoding)['all_weights']
B = all_data.shape[0]
seed = utilities.check_random_state(0)
shuffle = np.argsort(seed.rand(B))
train_data = all_data[shuffle][:int(0.8 * B)]
test_data = all_data[shuffle][int(0.8 * B):]
train_weights = all_weights[shuffle][:int(0.8 * B)]
test_weights = all_weights[shuffle][int(0.8 * B):]
local_env['train_data'] = train_data.astype(curr_int)
local_env['test_data'] = test_data.astype(curr_int)
local_env['train_weights'] = train_weights.astype(curr_float)
local_env['test_weights'] = test_weights.astype(curr_float)
def load_Kunitz(local_env, path=''):
all_data = Proteins_utils.load_FASTA(path + 'Kunitz_domain_MSA.fasta')
num_neighbours = Proteins_utils.count_neighbours(all_data)
all_weights = 1.0 / num_neighbours
B = all_data.shape[0]
seed = utilities.check_random_state(0)
shuffle = np.argsort(seed.rand(B))
train_data = all_data[shuffle][:int(0.8 * B)]
test_data = all_data[shuffle][int(0.8 * B):]
train_weights = all_weights[shuffle][:int(0.8 * B)]
test_weights = all_weights[shuffle][int(0.8 * B):]
from scipy.io import loadmat
contact_map = loadmat(path + 'contact_map14_extended.mat')['cm'] > 0
contact_map += contact_map.T # Load contact map.
local_env['train_data'] = train_data.astype(curr_int)
local_env['test_data'] = test_data.astype(curr_int)
local_env['train_weights'] = train_weights.astype(curr_float)
local_env['test_weights'] = test_weights.astype(curr_float)
local_env['contact_map'] = contact_map
def __init__(self, N=10, M=5, n_c=2, nature='Potts', gauge='zerosum', random_state=None):
self.N = N
self.M = M
self.n_c = n_c
self.gauge = gauge
self.nature = nature
self.muh = np.ones(M, dtype=curr_float) / M
self.cum_muh = np.cumsum(self.muh)
self.gh = np.zeros(M, dtype=curr_float)
if nature == 'Potts':
self.weights = np.zeros([M, N, n_c], dtype=curr_float)
else:
self.weights = np.zeros([M, N], dtype=curr_float)
if nature == 'Bernoulli':
self.cond_muv = np.ones([M, N], dtype=curr_float) / 2
elif nature == 'Spin':
self.cond_muv = np.zeros([M, N], dtype=curr_float)
elif nature == 'Potts':
self.cond_muv = np.ones([M, N, n_c], dtype=curr_float) / n_c
self.cum_cond_muv = np.cumsum(self.cond_muv, axis=-1)
self.random_state = utilities.check_random_state(random_state)
self.logpartition()
figsize=(h, w),
ticks_every=10,
ticks_labels_size=10,
title_size=12,
dpi=400,
molecule=molecule)
dest = './trial/'
g15p = './trial/v3_c0_all.txt'
# import data and weights
all_data = Proteins_utils.load_FASTA(g15p, drop_duplicates=True, type='dna')
affs = get_affinities(g15p, all_data)
seed = utilities.check_random_state(0)
permutation = np.argsort(seed.rand(all_data.shape[0]))
affs = affs[permutation]
all_data = all_data[permutation] # Shuffle data.
#WEIGHTS
num_neighbours = Proteins_utils.count_neighbours(all_data)
# all_weights = 1.0/num_neighbours
weights = np.asarray([float(i) / 1000. for i in affs], dtype='float')
#mu = utilities.average(all_data,c=4,weights=all_weights)
#sequence_logo.Sequence_logo(mu,ticks_every=5);
#PARAMETERS