def fit(self):
""" """
i = 0
errs_models = []
errs_null_models = []
nb_coefs = []
nb_models = []
intercepts = []
print '\n######## cross validation\n####parameters:{0}'\
.format(self.ssrge_params)
ssrge = SSrGE(**self.ssrge_params)
kfold = KFold(n_splits=self.n_folds, shuffle=True)
for train, test in kfold.split(self.SNV_mat):
i += 1
print '\n## fold nb {0}'.format(i)
X_train = self.SNV_mat[train]
Y_train = self.GE_mat.T[train].T
X_test = self.SNV_mat[test]
Y_test = self.GE_mat.T[test].T
ssrge.fit(X_train, Y_train)
score, score_null = ssrge.score(X_test, Y_test)
errs_models.append(score)
errs_null_models.append(score_null)
nb_coefs.append(len(ssrge.eeSNV_weight))
nb_models.append(len(ssrge.intercepts))
intercepts.append(np.mean(ssrge.intercepts.values()))
return (np.mean(errs_models), np.mean(errs_null_models),
np.mean(nb_coefs), np.mean(nb_models), np.mean(intercepts))
def test_ssrge_cnv(self):
"""test ssrge procedure with cnv matrix"""
from garmire_SSrGE.examples import create_example_matrix_v3
from garmire_SSrGE.ssrge import SSrGE
X, Y, C, W = create_example_matrix_v3()
ssrge = SSrGE(alpha=0.01)
ssrge.fit(X, Y, C)
self.assertTrue(ssrge.eeSNV_weight)
Xr = ssrge.transform(X)
self.assertTrue(Xr.sum())
self.assertTrue(Xr.shape[0] == X.shape[0])
self.assertTrue(Xr.shape[1] < X.shape[1])
snv_ranked = ssrge.rank_eeSNVs()
self.assertTrue(snv_ranked)
score = ssrge.score(X, Y)
self.assertTrue(score[0] < score[1])
def test_ssrge_rank_gene(self):
"""test ssrge and rank genes and snvs"""
from garmire_SSrGE.examples import create_example_matrix_v2
from garmire_SSrGE.ssrge import SSrGE
X, Y, gene_id_list, snv_id_list = create_example_matrix_v2()
ssrge = SSrGE(snv_id_list=snv_id_list,
gene_id_list=gene_id_list,
nb_ranked_features=2,
alpha=0.01)
ssrge.fit(X, Y)
self.assertTrue(ssrge.eeSNV_weight)
Xr = ssrge.transform(X)
self.assertTrue(Xr.sum())
self.assertTrue(Xr.shape[0] == X.shape[0])
self.assertTrue(Xr.shape[1] < X.shape[1])
snv_ranked = ssrge.rank_eeSNVs()
self.assertTrue(snv_ranked)
self.assertTrue(len(ssrge.retained_snvs) == ssrge.nb_ranked_features)
self.assertTrue(len(ssrge.retained_genes) == ssrge.nb_ranked_features)
score = ssrge.score(X, Y)
self.assertTrue(score[0] < score[1])
subgroup = ssrge.rank_features_for_a_subgroup(range(10))
self.assertTrue(len(subgroup.gene_expr_distrib[gene_id_list[0]]) == 10)
self.assertTrue(subgroup.snv_weights_distrib)
self.assertTrue(subgroup.exp_snv_distrib_comp)
def test_ssrge_elasticnet(self):
"""test ssrge procedure with elasticnet model"""
from garmire_SSrGE.examples import create_example_matrix_v1
from garmire_SSrGE.ssrge import SSrGE
X, Y, W = create_example_matrix_v1()
ssrge = SSrGE(alpha=0.01, model='ElasticNet')
ssrge.fit(X, Y)
self.assertTrue(ssrge.eeSNV_weight)
Xr = ssrge.transform(X)
self.assertTrue(Xr.sum())
self.assertTrue(Xr.shape[0] == X.shape[0])
self.assertTrue(Xr.shape[1] <= X.shape[1])
snv_ranked = ssrge.rank_eeSNVs()
self.assertTrue(snv_ranked)
score = ssrge.score(X, Y)
self.assertTrue(score[0] < score[1])