def _convertGaCall(self, recordId, name, pysamCall, genotypeData):
compoundId = self.getCallSetId(name)
callSet = self.getCallSet(compoundId)
call = protocol.Call()
call.callSetId = callSet.getId()
call.callSetName = callSet.getSampleName()
call.sampleId = callSet.getSampleName()
# TODO:
# NOTE: THE FOLLOWING TWO LINES IS NOT THE INTENDED IMPLEMENTATION,
###########################################
call.phaseset = None
call.genotype, call.phaseset = convertVCFGenotype(
genotypeData, call.phaseset)
###########################################
# THEY SHOULD BE REPLACED BY THE FOLLOWING, ONCE NEW PYSAM
# RELEASE SUPPORTS phaseset. AS WELL AS REMOVING genotypeData
# FROM THE FUNCTION CALL
###########################################
# call.genotype = list(pysamCall.allele_indices)
# call.phaseset = pysamCall.phaseset
###########################################
call.genotypeLikelihood = []
for key, value in pysamCall.iteritems():
if key == 'GL' and value is not None:
call.genotypeLikelihood = list(value)
elif key != 'GT':
call.info[key] = _encodeValue(value)
return call
def generateVariant(self, referenceName, position, randomNumberGenerator):
"""
Generate a random variant for the specified position using the
specified random number generator. This generator should be seeded
with a value that is unique to this position so that the same variant
will always be produced regardless of the order it is generated in.
"""
variant = self._createGaVariant()
variant.names = []
variant.referenceName = referenceName
variant.start = position
variant.end = position + 1 # SNPs only for now
bases = ["A", "C", "G", "T"]
ref = randomNumberGenerator.choice(bases)
variant.referenceBases = ref
alt = randomNumberGenerator.choice(
[base for base in bases if base != ref])
variant.alternateBases = [alt]
variant.calls = []
for callSet in self.getCallSets():
call = protocol.Call()
call.callSetId = callSet.getId()
# for now, the genotype is either [0,1], [1,1] or [1,0] with equal
# probability; probably will want to do something more
# sophisticated later.
randomChoice = randomNumberGenerator.choice(
[[0, 1], [1, 0], [1, 1]])
call.genotype = randomChoice
# TODO What is a reasonable model for generating these likelihoods?
# Are these log-scaled? Spec does not say.
call.genotypeLikelihood = [-100, -100, -100]
variant.calls.append(call)
variant.id = self.getVariantId(variant)
return variant