def clusterize_by_rt(self, error_rt):
"""
PUBLIC function
Provide a basic clustering home made
:param error_rt:
return: list of clusters(as set)
"""
if self.rt_method == 1:
logging.info("Basic clustering with rt_error:{0}".format(error_rt))
if not isinstance(error_rt, float):
raise TypeError("[clusterize]: args[0] is not a float")
return clusterize_basic(self.peakels, self.BASIC_RT_CALLABLE, error_rt)
elif self.rt_method == 2:
rts = [[x.rt] for x in self.peakels]
matrix_dist = sp.spatial.distance.pdist(np.array(rts)) # metric = eclidean by default
return list(clusterize_hierarchical(self.peakels, matrix_dist, "", error_rt).values())
elif self.rt_method == 3:
logging.info("DB SCAN clustering with error_rt:{0}".format(error_rt))
rts = [[x.rt] for x in self.peakels]
clusters = clusterize_dbscan(rts, self.peakels, eps=0.35)
# with open('clusters.txt', 'w') as f:
# for c in clusters:
# for fe in c:
# f.write(str(fe.rt) + '\n')
# f.write('\n')
return clusters # eps=error_rt / 2.0, min_samples=1)
else:
raise ValueError("wrong clustering technique !")
def clusterize_by_rt(self, error_rt):
"""
PUBLIC function
Provide a basic clustering home made
:param error_rt:
return: list of clusters(as set)
"""
if self.rt_method == 1:
logging.info("Basic clustering with rt_error:{0}".format(error_rt))
if not isinstance(error_rt, float):
raise TypeError("[clusterize]: args[0] is not a float")
return clusterize_basic(self.peakels, self.BASIC_RT_CALLABLE, error_rt)
elif self.rt_method == 2:
rts = [[x.rt] for x in self.peakels]
matrix_dist = sp.spatial.distance.pdist(np.array(rts)) # metric = eclidean by default
return list(clusterize_hierarchical(self.peakels, matrix_dist, "", error_rt).values())
elif self.rt_method == 3:
logging.info('DB SCAN clustering with error_rt:{0}'.format(error_rt))
rts = [[x.rt] for x in self.peakels]
clusters = clusterize_dbscan(rts, self.peakels, eps=0.35)
# with open('clusters.txt', 'w') as f:
# for c in clusters:
# for fe in c:
# f.write(str(fe.rt) + '\n')
# f.write('\n')
return clusters # eps=error_rt / 2.0, min_samples=1)
else:
raise ValueError("wrong clustering technique !")
def _check_update_corr_shape_in_rt_cluster(self, rt_cluster, distance_corr=DEFAULT_SHAPE_CORR):
"""
PRIVATE function
calculate corral
"""
clust_list = None
if self.corr_shape_method == 1:
clust_list = clusterize_basic(rt_cluster, self.BASIC_CORR_SHAPE_CALLABLE, distance_corr)
elif self.corr_shape_method == 2:
ints = [[y.intensity for y in x.peaks] if len(x.peaks) else [0] for x in rt_cluster]
matrix_dist = sp.spatial.distance.pdist(np.array(ints), metric="correlation")
clust_list = clusterize_hierarchical(rt_cluster, matrix_dist, distance_corr, clip=True)
return self._split_rt_cluster(clust_list)
def _check_update_corr_shape_in_rt_cluster(self, rt_cluster, distance_corr=DEFAULT_SHAPE_CORR):
"""
PRIVATE function
calculate corral
"""
clust_list = None
if self.corr_shape_method == 1:
clust_list = clusterize_basic(rt_cluster, self.BASIC_CORR_SHAPE_CALLABLE, distance_corr)
elif self.corr_shape_method == 2:
ints = [[y.intensity for y in x.peaks] if len(x.peaks) else [0] for x in rt_cluster]
matrix_dist = sp.spatial.distance.pdist(np.array(ints), metric='correlation')
clust_list = clusterize_hierarchical(rt_cluster, matrix_dist, distance_corr, clip=True)
return self._split_rt_cluster(clust_list)
def _check_update_corr_intensity_in_rt_cluster(self, rt_cluster, distance_corr=DEFAULT_INT_CORR):
"""
Private function
"""
if len(rt_cluster) == 1:
return [] # rt_cluster, []
# clust_list = None
if self.corr_int_method == 1:
clust_list = clusterize_basic(rt_cluster, self.BASIC_CORR_INT_CALLABLE, distance_corr)
elif self.corr_int_method == 2:
ints = [list(x.area_by_sample_name.values()) for x in rt_cluster] #
# matrix_dist = sp.spatial.distance.pdist(np.array(ints), metric='correlation')
# ude by default all cores on the machine
matrix_dist = pairwise_distances(np.array(ints), metric="correlation") # , n_jobs=-1)
clust_list = clusterize_hierarchical(rt_cluster, matrix_dist, distance_corr, clip=True)
else:
raise ValueError("dbscan not supported for intensities correlation clustering")
return clust_list # self._split_rt_cluster(clust_list)
def _check_update_corr_intensity_in_rt_cluster(self, rt_cluster, distance_corr=DEFAULT_INT_CORR):
"""
Private function
"""
if len(rt_cluster) == 1:
return [] # rt_cluster, []
# clust_list = None
if self.corr_int_method == 1:
clust_list = clusterize_basic(rt_cluster, self.BASIC_CORR_INT_CALLABLE, distance_corr)
elif self.corr_int_method == 2:
ints = [list(x.area_by_sample_name.values()) for x in rt_cluster] #
# matrix_dist = sp.spatial.distance.pdist(np.array(ints), metric='correlation')
# ude by default all cores on the machine
matrix_dist = pairwise_distances(np.array(ints), metric='correlation') # , n_jobs=-1)
clust_list = clusterize_hierarchical(rt_cluster, matrix_dist, distance_corr, clip=True)
else:
raise ValueError("dbscan not supported for intensities correlation clustering")
return clust_list # self._split_rt_cluster(clust_list)
def test_clusterize_hierarchical_int(self):
ints = [list(f.area_by_sample_name.values()) for f in self.features]
matrix_dist = sp.spatial.distance.pdist(np.array(ints), metric="correlation") # euclidean distance
clusters = clusterize_hierarchical(self.features, matrix_dist, 0.1, clip=True)
print(("len clusters hierarchical int: {0}".format(len(clusters))))
self.assertGreaterEqual(4, len(clusters))
def test_clusterize_hierarchical(self):
rts = [[f.rt] for f in self.features]
matrix_dist = sp.spatial.distance.pdist(np.array(rts)) # euclidean distance
clusters = clusterize_hierarchical(self.features, matrix_dist, 3.0)
print(("len clusters hierarchical: {0}".format(len(clusters))))
self.assertGreaterEqual(4, len(clusters))