def train_tissue_specific_genome_only(self):
for i in range(self.num_tissues):
# train tissue-specific model
beta = lr.sgd(self.train_list[i][self.genomic_features].values,
self.train_list[i]["expr_label"].values,
np.zeros(len(self.genomic_features)),
np.zeros(len(self.genomic_features)), 1.0)
self.train_list[i]["tissue specific genome only"] = np.exp(
lr.log_prob(self.train_list[i][self.genomic_features].values,
beta))
self.test_list[i]["tissue specific genome only"] = np.exp(
lr.log_prob(self.test_list[i][self.genomic_features].values,
beta))
def train_shared_tissue_genome_only(self):
beta = lr.sgd(self.train_list[0][self.genomic_features].values,
self.train_list[0]["median_expr_label"].values,
np.zeros(len(self.genomic_features)),
np.zeros(len(self.genomic_features)), 1.0)
for i in range(self.num_tissues):
self.train_list[i]["shared tissue genome only"] = np.exp(
lr.log_prob(self.train_list[i][self.genomic_features].values,
beta))
self.test_list[i]["shared tissue genome only"] = np.exp(
lr.log_prob(self.test_list[i][self.genomic_features].values,
beta))
def computeLikelihood(self):
ll = self.log_p_beta()
# P(beta^c | beta)
for i in range(self.num_tissues):
ll += self.log_p_beta_child_given_beta(i)
for i in range(self.num_tissues):
try:
log_prob_z_1_g = lr.log_prob(
self.train_list[i][self.genomic_features], self.getBetaLeaf(i))
log_prob_z_0_g = np.log(1.0 - np.exp(log_prob_z_1_g))
log_prob_e_z_1 = nb.log_prob(self.train_list[i]['expr_label'], 1,
self.phi)
b = log_prob_e_z_1 + lr.log_prob(
self.train_list[i][self.genomic_features], self.getBetaLeaf(i))
except:
continue
a = nb.log_prob(self.train_list[i]['expr_label'], 0,
self.phi) + np.log(1.0 - np.exp(log_prob_z_1_g))
# log sum exp trick
s = np.maximum(a, b)
unnormalized_prob = s + np.log(np.exp(a - s) + np.exp(b - s))
ll_tissue = np.nansum(unnormalized_prob)
ll += ll_tissue
def _RIVER_likelihood(e, g, beta, phi):
# log p(z = 1 | g)
log_p_z_1_given_g = lr.log_prob(g, beta)
# log p(z = 0 | g)
log_p_z_0_given_g = np.log(1.0 - np.exp(log_p_z_1_given_g))
# log p(e | z = 1)
log_p_e_given_z_1 = nb.log_prob(e, 1, phi)
# log p(e | z = 0)
log_p_e_given_z_0 = nb.log_prob(e, 0, phi)
import pdb
pdb.set_trace()
#x_1 =
#m = np.maximum()
return 1
def _cross_validate(self, G, E):
'''
K-fold Cross-validate beta MAP estimation to find optimal lambda
:param G genomic features
:param E expression labels
:lambda set
'''
G = G
E = E
# lambda set
lambda_set = np.array([
1e-6, 1e-5, 1e-4, 1e-3, 1e-2, 1e-1, 1e0, 1e1, 1e2, 1e3, 1e4, 1e5,
1e6
])
# initialize beta to zero
beta_init = np.zeros(len(self.genomic_features))
# AUC scores for each lambda and each fold
scores_list = np.zeros((len(lambda_set), self.num_folds))
# for each fold
for k in range(self.num_folds):
# access training data (everything but Kth fold)
training = np.array(
[x for i, x in enumerate(G) if i % self.num_folds != k])
training_labels = np.array(
[x for i, x in enumerate(E) if i % self.num_folds != k])
# access validation data (Kth fold)
validation = np.array(
[[x for i, x in enumerate(G) if i % self.num_folds == k]])
validation_labels = np.array(
[x for i, x in enumerate(E) if i % self.num_folds == k])
# for each possible lambda
for i in range(len(lambda_set)):
# train a logistic regression model
beta = lr.sgd(training, training_labels, beta_init, beta_init,
float(lambda_set[i]))
# compute predictions on validation set
scores = lr.log_prob(validation, beta).reshape(-1)
# compute auc using predictions and validation set labels
auc = sklearn.metrics.roc_auc_score(validation_labels, scores)
scores_list[i][k] = auc
# average across all folds for each lambda
lambda_averages = np.mean(scores_list, axis=1)
# sanity check
assert len(lambda_averages) == len(lambda_set)
optimal_lambda = lambda_set[np.argmax(lambda_averages)]
return optimal_lambda
def eStepLocal(self, i, data, beta, phi):
'''
Compute p(z | ...) for tissue i
i : int
tissue index
data : panda data frame
core data structure containing genomic features, expression, updated posteriors.
beta : numpy array : 1 x M
coefficients for genomic features
phi : numpy array
either 2 x 2 numpy array for categorical distribution or 1 x 2 for noisy or
'''
# log p(z | g)
log_prob_z_1_given_g = lr.log_prob(data[self.genomic_features].values,
beta)
log_prob_z_0_given_g = np.log(1.0 - np.exp(log_prob_z_1_given_g))
# log p(e | z, q)
if self.e_distribution == 'noisyor':
# noisy OR
log_prob_e_given_z_1 = nb.log_prob_noisyor_2_params(
data['expr_label'], 1, data["eqtl"], phi)
log_prob_e_given_z_0 = nb.log_prob_noisyor_2_params(
data[i]['expr_label'], 0, data["eqtl"], phi)
# log p(e | z)
else:
# naive bayes
log_prob_e_given_z_1 = nb.log_prob(data['expr_label'].values, 1,
self.phi)
log_prob_e_given_z_0 = nb.log_prob(data['expr_label'].values, 0,
self.phi)
# p(e|z =1) * p(z = 1 | g) / (\sum_{z \in S} p(z = s | g) * p(e | z = s))
log_q = log_prob_e_given_z_1 + log_prob_z_1_given_g - np.log(
np.exp(log_prob_e_given_z_0) * np.exp(log_prob_z_0_given_g) +
np.exp(log_prob_e_given_z_1) * np.exp(log_prob_z_1_given_g))
return np.exp(log_q)
def eStepLocalTest(self, i, beta, phi):
'''
Compute expectation for tissue i
'''
# log P(Z = 1 | G)
log_prob_z_1_given_g = lr.log_prob(
self.test_list[i][self.genomic_features].values, beta)
# log P(Z = 0 | G)
log_prob_z_0_given_g = np.log(1.0 - np.exp(log_prob_z_1_given_g))
# log P(E | Z = 1)
log_prob_e_given_z_1 = nb.log_prob(
self.test_list[i][self.label].values, 1, phi)
# log P(E | Z = 0)
log_prob_e_given_z_0 = nb.log_prob(
self.test_list[i][self.label].values, 0, phi)
log_q = log_prob_e_given_z_1 + log_prob_z_1_given_g - np.log(
np.exp(log_prob_e_given_z_0) * np.exp(log_prob_z_0_given_g) +
np.exp(log_prob_e_given_z_1) * np.exp(log_prob_z_1_given_g))
return np.exp(log_q)
def _compute_p_z_given_g(self, beta, g): ''' P(z | g; beta) ''' return np.exp(lr.log_prob(g, beta))