def dotrueiteration(seqsdata, W, pwm, lambda_):
"""Do one true iteration of EM. I.e. iterate over all W-mers."""
logger.debug('Calculating true Zn sums')
summer = jis.ZnSumCb(W)
calculateZn = jem.createZncalculatorFn(pwm, lambda_)
sumvisitor = jis.ZnCalcVisitor(W, calculateZn, summer)
seqan.traverse.topdownhistorytraversal(
seqsdata.index.topdownhistory(), sumvisitor)
return summer
def dotrueiteration(seqsdata, W, pwm, lambda_):
"""Do one true iteration of EM. I.e. iterate over all W-mers."""
logger.debug('Calculating true Zn sums')
summer = jis.ZnSumCb(W)
calculateZn = jem.createZncalculatorFn(pwm, lambda_)
sumvisitor = jis.ZnCalcVisitor(W, calculateZn, summer)
seqan.traverse.topdownhistorytraversal(seqsdata.index.topdownhistory(),
sumvisitor)
return summer
def uniformuniquemethod(seqsdata, pwm, lambda_, Widx, numsamples, args):
logger.debug('Importance sampling using uniform weights')
calculateZn = jem.createZncalculatorFn(pwm, lambda_)
W = args.Ws[Widx]
childuniquefreqs = jis.DistForFreqs(seqsdata.childuniquefreqs[:, Widx])
return seqsdata.numunique[Widx], jis.importancesample(
seqsdata.index, W,
jis.WeightedSamplingDist(childuniquefreqs,
jis.UniformImportanceWeight()), numsamples,
jis.ISMemoCbAdaptor(W, jis.ZnSumCb(W), calculateZn, unique=True))
def uniformuniquemethod(seqsdata, pwm, lambda_, Widx, numsamples, args):
logger.debug('Importance sampling using uniform weights')
calculateZn = jem.createZncalculatorFn(pwm, lambda_)
W = args.Ws[Widx]
childuniquefreqs = jis.DistForFreqs(seqsdata.childuniquefreqs[:, Widx])
return seqsdata.numunique[Widx], jis.importancesample(
seqsdata.index,
W,
jis.WeightedSamplingDist(
childuniquefreqs, jis.UniformImportanceWeight()),
numsamples,
jis.ISMemoCbAdaptor(W, jis.ZnSumCb(W), calculateZn, unique=True))
def pwmoccsmethod(seqsdata, pwm, lambda_, Widx, numsamples, args):
logger.debug('Importance sampling using PWM importance weights')
calculateZn = jem.createZncalculatorFn(pwm, lambda_)
numpositive = numsamples / 2 # Sample half in each orientation
W = args.Ws[Widx]
childoccfreqs = jis.DistForFreqs(seqsdata.childoccfreqs[:, Widx])
cb = jis.importancesample(
seqsdata.index, W,
jis.WeightedSamplingDist(childoccfreqs, jis.PWMImportanceWeight(pwm)),
numpositive, jis.ISMemoCbAdaptor(W, jis.ZnSumCb(W), calculateZn))
return seqsdata.numoccs[Widx], jis.importancesample(
seqsdata.index, W,
jis.WeightedSamplingDist(childoccfreqs,
jis.PWMImportanceWeight(jem.pwmrevcomp(pwm))),
numsamples - numpositive, cb)
def pwmuniquemethod(seqsdata, pwm, lambda_, Widx, numsamples, args):
logger.debug('Importance sampling using PWM weights over unique W-mers')
calculateZn = jem.createZncalculatorFn(pwm, lambda_)
W = args.Ws[Widx]
childuniquefreqs = jis.DistForFreqs(seqsdata.childuniquefreqs[:, Widx])
childoccfreqs = jis.DistForFreqs(seqsdata.childoccfreqs[:, Widx])
numpositive = numsamples / 2 # Sample half in each orientation
cb = jis.importancesample(
seqsdata.index, W,
jis.WeightedSamplingDist(
childuniquefreqs,
jis.PWMImportanceWeightUnique(pwm, childoccfreqs)), numpositive,
jis.ISMemoCbAdaptor(W, jis.ZnSumCb(W), calculateZn, unique=True))
return seqsdata.numunique[Widx], jis.importancesample(
seqsdata.index, W,
jis.WeightedSamplingDist(
childuniquefreqs,
jis.PWMImportanceWeightUnique(jem.pwmrevcomp(pwm), childoccfreqs)),
numsamples - numpositive, cb)
def pwmoccsmethod(seqsdata, pwm, lambda_, Widx, numsamples, args):
logger.debug('Importance sampling using PWM importance weights')
calculateZn = jem.createZncalculatorFn(pwm, lambda_)
numpositive = numsamples / 2 # Sample half in each orientation
W = args.Ws[Widx]
childoccfreqs = jis.DistForFreqs(seqsdata.childoccfreqs[:, Widx])
cb = jis.importancesample(
seqsdata.index,
W,
jis.WeightedSamplingDist(
childoccfreqs,
jis.PWMImportanceWeight(pwm)),
numpositive,
jis.ISMemoCbAdaptor(W, jis.ZnSumCb(W), calculateZn))
return seqsdata.numoccs[Widx], jis.importancesample(
seqsdata.index,
W,
jis.WeightedSamplingDist(
childoccfreqs,
jis.PWMImportanceWeight(jem.pwmrevcomp(pwm))),
numsamples - numpositive,
cb)
def pwmuniquemethod(seqsdata, pwm, lambda_, Widx, numsamples, args):
logger.debug('Importance sampling using PWM weights over unique W-mers')
calculateZn = jem.createZncalculatorFn(pwm, lambda_)
W = args.Ws[Widx]
childuniquefreqs = jis.DistForFreqs(seqsdata.childuniquefreqs[:, Widx])
childoccfreqs = jis.DistForFreqs(seqsdata.childoccfreqs[:, Widx])
numpositive = numsamples / 2 # Sample half in each orientation
cb = jis.importancesample(
seqsdata.index,
W,
jis.WeightedSamplingDist(
childuniquefreqs,
jis.PWMImportanceWeightUnique(pwm, childoccfreqs)),
numpositive,
jis.ISMemoCbAdaptor(W, jis.ZnSumCb(W), calculateZn, unique=True))
return seqsdata.numunique[Widx], jis.importancesample(
seqsdata.index,
W,
jis.WeightedSamplingDist(
childuniquefreqs,
jis.PWMImportanceWeightUnique(jem.pwmrevcomp(pwm), childoccfreqs)),
numsamples - numpositive,
cb)
memocb = jis.importancesample(
index, W, childWmerfreqs[:, 0], jis.UniformImportanceWeight(),
numsamples=1, callback=jis.ISCbMemo())
pwm = jem.pwmfromWmer(memocb.Xns[0], numseedsites, 1.)
jem.logo(pwm, 'seed')
numsamples = 3000
distsbs = []
distsbg = []
truesums = []
varratios = []
pwmtrue = pwm.copy()
for iteration in xrange(5):
logging.debug('Calculating true Zn sums')
summer = jis.ZnSumCb(W)
calculateZn = jem.createZncalculatorFn(pwmtrue, lambda_)
sumvisitor = jis.ZnCalcVisitor(W, calculateZn, summer)
seqan.traverse.topdownhistorytraversal(index.topdownhistory(), sumvisitor)
logging.debug('Sums:\n%s', summer.sums)
trueZnsum = summer.sums[0].sum()
truesums.append(trueZnsum)
logging.info('True sum: %s', trueZnsum)
logging.debug('Importance sampling using binding site model')
cbbs = jis.importancesample(
index, W, childWmerfreqs[:, 0], jis.PWMImportanceWeight(pwmtrue),
numsamples, jis.ISMemoCbAdaptor(jis.ZnSumCb(W), calculateZn))
samplebs = makedf(cbbs, calculateZn, model='BS')
logging.debug('Variances:\n%s', cbbs.cb.variances())
logging.debug('Importance sampling using background model')
