def _getUsable(self, comp=None, dose=None):
try:
f=open(self.settings.ok_wells_asDICT, 'r')
ok_dict=pickle.load(f);f.close()
except:
if comp is None:
return quality_control.usable_XBSC(self.compound, self.dose,confDir=self.settings.plate_setups_folder)
else:
return quality_control.usable_XBSC(comp, dose,confDir=self.settings.plate_setups_folder)
else:
if comp is None:
return ok_dict[self.compound][self.dose]
else:
return ok_dict[comp][dose]
def parameterStabilityEvaluation(well_num=50,n_list=list([0.4,0.5,0.6,0.7,0.8]), h_list=[10], deg_list=[2,3],
pheno_list=['Anaphase_ch1', 'Apoptosis_ch1', 'Folded_ch1', 'Interphase_ch1', 'Metaphase_ch1',
'Polylobbed_ch1', 'Prometaphase_ch1', 'WMicronuclei_ch1'],
loadingFolder='/media/lalil0u/New/projects/Xb_screen/dry_lab_results', passed=None):
'''
Here we're looking at how well results are robust to parameters change (result r1) and how well a subset of
parameters is able to separate experiments (variance of r2).
'''
param_num=len(n_list)*len(h_list)*len(deg_list)
param_list=[]
if passed is None:
passed,_,_,_=quality_control.computingToRedo(threshold_flou=0.4, threshold_init_cell_count=20)
selected=np.array(passed)[np.random.permutation(len(passed))[:well_num]]
processedDict={};
result=defaultdict(list)
for pl,w in selected:
if pl not in processedDict:
print "Opening processed dictionary file for pl ", pl
f=open(os.path.join(loadingFolder, "processedDictResult_P{}.pkl".format(pl)))
processedDict[pl]=pickle.load(f); f.close()
try:
currCtrl=quality_control.usable_XBSC(CONTROLS[processedDict[pl][int(w)]['Xenobiotic']], 0, pl)
except KeyError:
if processedDict[pl][int(w)]['Xenobiotic'] in CONTROLS.values():
currCtrl=quality_control.usable_XBSC(processedDict[pl][int(w)]['Xenobiotic'], 0, pl)
else:
raise KeyError
for pheno in pheno_list:
y_exp=np.array(processedDict[pl][int(w)][pheno])
for n,h,deg in product(n_list, h_list, deg_list):
if (n,h,deg) not in param_list:
param_list.append((n,h,deg))
r=[]
for pl, ctrl in filter(lambda x: x[1]!=w,currCtrl):
y_ctrl = np.array(processedDict[pl][int(ctrl)][pheno])
_, r_exp=localReg(y_exp, n, h, deg, plot=False)
_, r_ctrl = localReg(y_ctrl, n, h, deg, plot=False)
r.append(np.max(np.abs(r_exp-r_ctrl)))
result[pheno].append(np.median(r))
print "Done with local regressions"
for pheno in pheno_list:
np.reshape(result[pheno], newshape=(well_num, param_num))
r1=[np.ones(shape=(param_num, param_num), dtype=float) for pheno in pheno_list]
r2=[np.zeros(shape=(param_num,), dtype=list) for pheno in pheno_list]
for i, pheno in enumerate(pheno_list):
result[pheno]=np.reshape(result[pheno], newshape=(well_num, param_num))
for k in range(param_num):
r2[i][k]=[0]
iter_ = np.nditer(result[pheno][1:,k], flags=['c_index']) #one-dim array: the order is the same in C or F
while not iter_.finished :
r2[i][k]= np.hstack((r2[i][k],[a-iter_[0] for a in np.nditer(result[pheno][:iter_.index+1,k])]))
iter_.iternext()
r2[i][k]=np.abs(r2[i][k][1:])/np.mean(result[pheno][:,k])
for j in range(k):
r1[i][k,j]=spearmanr(result[pheno][:,k],result[pheno][:,j])[0]
r1[i][j,k]=r1[i][k,j]
return passed, param_list,r1, r2