When I run the following:

bcftools mpileup -BI -Ov -d 400000 -a AD,DP -o sample.A.vcf -f ref.fasta -R regions.bed sample.A.sorted.bam

I get the following output at a specific position:

#CHROM  POS ID  REF ALT QUAL    FILTER  INFO    FORMAT  sampleA
contig00012 439 .   C   A,<*>   0   .   DP=1365;I16=324,244,246,260,27639,1.51696e+06,24253,1.30823e+06,69,1683,106,4162,12757,303189,10513,247245;QS=0.523724,0.476276,0;VDB=0.473402;SGB=-0.693147;RPB=0.000285419;MQB=0.999823;MQSB=0.99981;BQB=0.024168;MQ0F=0.782418   PL:DP:AD    0,12,0,255,255,1:1074:568,506,0

Which from the AD field indicates that there are 568 alleles supporting the reference C allele and 506 supporting the alternative A allele.

However when I pass this to bcftools to call genotypes using:

bcftools call -mO v -o sample_A_called.vcf sample_A.vcf

I get the following:

#CHROM  POS ID  REF ALT QUAL    FILTER  INFO    FORMAT  DB31-92
contig00012 439 .   C   .   25.9297 .   DP=1365;VDB=0.473402;SGB=-0.693147;RPB=0.000285419;MQB=0.999823;MQSB=0.99981;BQB=0.024168;MQ0F=0.782418;AN=2;DP4=324,244,246,260;MQ=0   GT:DP:AD    0/0:1074:568

where it is now called homozygous reference (0/0) and the allelic depth for the alternative A allele seems to be missing.

From a quick visual inspection of the bam file this appears to be a true variant position and the base quality appears to be good.

Any suggestions on the reason for this behavior is greatly appreciated. Many thanks.