def _call_is_hom(variations, is_missing=None):
gts = variations[GT_FIELD]
is_hom = va.map_blocks(_call_is_hom_in_memory, gts, drop_axis=2)
if is_missing is not None:
is_hom[is_missing] = False
return is_hom
def count_alleles(gts, max_alleles, count_missing=True):
def _count_alleles(gts):
return _count_alleles_in_memory(gts, max_alleles, count_missing=count_missing)
try:
chunks = va.reduce_chunk_dimensions(gts)
except IndexError:
raise EmptyVariationsError()
allele_counts_by_snp = va.map_blocks(_count_alleles, gts, chunks=chunks,
drop_axis=(2,))
return allele_counts_by_snp
def calc_mac(variations, max_alleles,
min_num_genotypes=MIN_NUM_GENOTYPES_FOR_POP_STAT):
gts = variations[GT_FIELD]
# determine output chunks - preserve axis0; change axis1, axis2
# chunks = (gts.chunks[0])
chunks = None
def _private_calc_mac(gts):
return _calc_mac(gts, max_alleles=max_alleles)
macs = va.map_blocks(_private_calc_mac, gts, chunks=chunks,
drop_axis=(1, 2), dtype=numpy.float64)
return _mask_stats_with_few_samples(macs, variations, min_num_genotypes)
def _gt_to_missing(variations, field, min_value):
gts = variations[GT_FIELD]
calls_setted_to_missing = variations[field] < min_value
axis = 2
def _stack_in_memory(array):
return stack_in_memory(array, axis=axis)
# as we can not slice using arrays of diferente dimensions, we need to
# create one with same dimensions with stack
p2 = va.map_blocks(_stack_in_memory,
calls_setted_to_missing,
dtype='i4',
new_axis=2)
gts[p2] = MISSING_INT
# va.assign_with_masking_value(gts, MISSING_INT, p2)
variations[GT_FIELD] = gts
return {FLT_VARS: variations}