def __init__(self,seq=''):
self.seq = seq
self.info = ''
self.logP = 0
self.c_wmerbgs= None
if seq:
self.c_intA = MDsupport.seq2int(seq)
self.compute_logP()
def EM_Cstart(self):
verbose = self.verbose
if verbose:
print "Seeding models..."
sys.stdout.flush()
self.seed_models()
#Initialize parameters
if not self.param.has_key('gamma'): self.param['gamma'] = 0.2
timings = {'Probes':0, 'Background':0, 'C EM':0, 'Post':0}
_time = time.time()
for seq in self.seqs:
P = Probe(seq)
self.probes.append(P)
_time2 = time.time(); timings['Probes'] = _time2-_time; _time = _time2
if verbose: print "Optimizing candidates by EM."
if verbose: sys.stdout.flush()
c_logZ_sets = {}
for Model,i in zip(self.models,range(len(self.models))):
width = Model.width
self.calcmask(width)
if not c_logZ_sets.has_key(width):
c_logZs_set = []
if verbose: print "#%s |%s|"%(' '*28,'-'*len(self.seqs))
if verbose: sys.stdout.flush()
if verbose: print "Computing background (width %2d) "%width,
for P in self.probes:
if verbose: sys.stdout.write('.')
if verbose: sys.stdout.flush()
logZs = self.all_Wmers(width,P)
c_logZs = MDsupport.list2double(logZs)
c_logZs_set.append(c_logZs)
#P.c_wmerbgs = MDsupport.list2double(map(lambda x: x.logQtot, Wlist))
c_logZ_sets[width] = c_logZs_set
if verbose: print
c_logZ_set = c_logZ_sets[width]
for P,c_logZs in zip(self.probes,c_logZ_set):
P.c_wmerbgs = c_logZs
_time2 = time.time()
timings['Background'] = timings['Background'] +_time2-_time
_time = _time2
'''Perform EM'''
_time = time.time()
newModel = self.EM_C(Model, self.probes)
_time2 = time.time(); timings['C EM'] = timings['C EM'] + _time2-_time; _time = _time2
#print "cLL: ",newModel.joint
#print "pLL: ",self.compute_joint(newModel,Wmers_by_seq)
'''Was there a problem?'''
if newModel == None:
continue
'''Set various things in PSSM'''
#Distance(s)
seeddist = MotifTools.infomaskdiff(newModel,Model)
print '%s ----> %s'%(Model,newModel)
print "Seed %2d: %s --> %s mask:%9.5f infoMask:%9.5f d:%9.5f"%(
i, Model, newModel,
MotifTools.maskdiff(newModel,Model),
MotifTools.infomaskdiff(newModel,Model), #order is important
Model-newModel)
#Seed
if Model.seedtxt: newModel.seedtxt = Model.seedtxt
if Model.source: newModel.source = Model.source
#newModel.denoise()
newModel.seeddist = seeddist
newModel.seednum = i
print newModel
newModel._print_p()
newModel._print_ll()
'''Set various things in Candidate (like a wrapper for PSSM)'''
C = MotifCandidate()
C.pssm = newModel.copy()
#C.wmers = self.best_by_Z(Wmers_by_seq)
C.wmers = [newModel.emit() for junk in range(20)]
#C._update() #MAJOR REMOVAL????????? DBG 10-14-03
#C.MAPpurge()
C.pssm = newModel.copy()
self.candidates.append(C)
_time2 = time.time(); timings['Post'] = timings['Post']+_time2-_time;_time = _time2
'''Print Timing Information'''
if verbose:
print "# Timing Information"
_t = 0
for timing in timings.keys():
_t = _t + timings[timing]
for timing in timings.keys():
print "# %12s %f %f%%"%(timing,timings[timing],timings[timing]*100/_t)