def bayes_iteration(self):
logging.info("Iteratioin...")
PPI_choice = [False, True]
for PPI in PPI_choice:
self = pickle.load(
open("/%s/%s_self_constant_optimization.p" % (self.output_directory, self.output_prefix), "rb")
)
self.PPI_prior = PPI
if self.PPI_diagnosis == False and self.PPI_prior == True:
pass
else:
self.outliers_removal = False
for (tgid, tginst) in self.allgenedict.iteritems():
iteratenbem(
tginst,
debug=False,
meanvarmodel=self.mrm,
restart=False,
PPI_prior=self.PPI_prior,
removeoutliers=self.outliers_removal,
size_factor=self.size_f,
beta1_prior_var=self.non_PPI_beta_prior_variance,
)
pickle.dump(
self,
open(
"/%s/%s_self_bayes_iteration_PPI_%s_outliers_removal_%s.p"
% (self.output_directory, self.output_prefix, self.PPI_prior, self.outliers_removal),
"wb",
),
)
# base_index=np.where(~self.design_matrix[:,1:].any(axis=1))[0]
# transform_list=[1 if i in base_index else -1 for i in range(self.design_matrix.shape[0])]
# transform_matrix=np.matrix(transform_list)
for (tgid, tginst) in self.allgenedict.items():
bayes_selection(tginst, log_list=self.log_list, selection_constant=self.selection_constant)
self.outliers_removal = True
for (tgid, tginst) in self.allgenedict.iteritems():
iteratenbem(
tginst,
debug=False,
meanvarmodel=self.mrm,
restart=False,
PPI_prior=self.PPI_prior,
removeoutliers=self.outliers_removal,
size_factor=self.size_f,
beta1_prior_var=self.non_PPI_beta_prior_variance,
)
pickle.dump(
self,
open(
"/%s/%s_self_bayes_iteration_PPI_%s_outliers_removal_%s.p"
% (self.output_directory, self.output_prefix, self.PPI_prior, self.outliers_removal),
"wb",
),
)
def bayes_init(self):
maxgene = np.inf
if self.negative_control == None:
self.allgenedict, self.invalid_gRNA_dict = read_gene_from_file(
self.count_table, includesamples=self.include_samples, negative_control=self.negative_control
)
else:
self.allgenedict, self.invalid_gRNA_dict, self.negative_control_gRNAs = read_gene_from_file(
self.count_table, includesamples=self.include_samples, negative_control=self.negative_control
)
# calculate the size factor
cttab_sel = {}
for (geneid, gk) in self.allgenedict.iteritems():
sgid = gk.sgrnaid
sgreadmat = gk.nb_count.getT().tolist()
for i in range(len(sgid)):
cttab_sel[sgid[i]] = sgreadmat[i]
if hasattr(self, "norm_method"):
if self.norm_method != "none":
self.size_f = normalizeCounts(
cttab_sel,
method=self.norm_method,
returnfactor=True,
reversefactor=True,
negative_control_gRNAs=self.negative_control_gRNAs,
)
else:
self.size_f = None
else:
self.size_f = normalizeCounts(cttab_sel, returnfactor=True, reversefactor=True)
logging.info("size factor: " + ",".join([str(x) for x in self.size_f]))
# logging.info(self.design_matrix)
desmat = self.design_matrix
for (tgid, tginst) in self.allgenedict.iteritems():
# if tgid=="INO80B":
tginst.design_mat = desmat
iteratenbem(tginst, debug=False, estimateeff=False, alpha_val=0.05, size_factor=self.size_f)
tginst.w_estimate = []
deviation(self.allgenedict)
self.non_PPI_beta_prior_variance = beta_non_PPI_prior_calculation(self.allgenedict)
pickle.dump(self, open("/%s/%s_self_bayes_init.p" % (self.output_directory, self.output_prefix), "wb"))
def bayes_major(self):
self = pickle.load(open("/%s/%s_self_fitting.p" % (self.output_directory, self.output_prefix), "rb"))
logging.info("Run the algorithm for the second time ...")
ngene = 0
for (tgid, tginst) in self.allgenedict.iteritems():
# logging.info(tgid)
iteratenbem(
tginst,
debug=False,
meanvarmodel=self.mrm,
restart=False,
size_factor=self.size_f,
beta1_prior_var=self.non_PPI_beta_prior_variance,
)
ngene += 1
for (tgid, tginst) in self.allgenedict.iteritems():
if len(tginst.w_estimate) == 0:
tginst.w_estimate = np.ones(len(tginst.sgrnaid))
pickle.dump(self, open("/%s/%s_self_bayes_major.p" % (self.output_directory, self.output_prefix), "wb"))