VarScan output different when choosing VCF and variant missing?

Question

[Pascal]

Hi,

I am using VarScan version 2.3.7 for variant calling in my tumour samples. Depending if I choose VCF as output I seem to get different results. Here is an example on a specific position where I am expecting to find one somatic mutation, but get none in both cases:

java -jar VarScan.v2.3.7.jar somatic <(samtools mpileup -l region.bed -q 1 -B -f GRCh37-lite.fa normal.bam  tumour.bam) output.txt --mpileup 1

Result is:

681 positions in mpileup file
366 had sufficient coverage for comparison
365 were called Reference
0 were mixed SNP-indel calls and filtered
1 were removed by the strand filter #Note this line
0 were called Germline
0 were called LOH
0 were called Somatic
0 were called Unknown
0 were called Variant

Same for VCF output:

java -jar VarScan.v2.3.7.jar somatic <(samtools mpileup -l region.bed -q 1 -B -f GRCh37-lite.fa normal.bam  tumour.bam) output.txt --mpileup 1 --output-vcf

Result is:

681 positions in mpileup file
366 had sufficient coverage for comparison
365 were called Reference
0 were mixed SNP-indel calls and filtered
0 were removed by the strand filter
0 were called Germline
0 were called LOH
1 were called Somatic #Note this line
0 were called Unknown
0 were called Variant

However, the result files are empty in both cases (apart from the header).

Help is highly appreciated.

Pascal

[Pascal]

Just tried version 2.3.1. With both options 1 somatic mutation is reported and both output files contain this mutation.

[sginzel]

Did you manage find the reason for this behaviour? Would be interested to how that might happen.

[Pascal]

No, I wasn't able to track down what happens there.