2.) Improved parameter-handling logic for two flags (--validation and --strand-filter), which previously were
sometimes considered "turned on" if the user provided them, even if the value provided was a zero.
This means that even though the default is to have the strand filter turned off (--strand-filter 0), it may get applied in prior versions, with variants "strand filtered" erroneously. This was "fixed" in 2.4.0. Nonetheless, we found that this "fix" gets rolled back in 2.4.1 and 2.4.2, and then re-applied in 2.4.3. Likely this fix got rolled back along with the changes with "the way normal_reads2 are counted". From the 2.4.0 release notes :
1.) A correction in the way normal_reads2 values are counted when the mutation allele is not observed. Prior
to this fix, a non-reference base would be counted as a variant allele even if it didn't match the actual mutation
allele called in the tumor. Now, only observations of the tumor variant allele will be counted and go into the FET.
Nonetheless, we found a variant that is strand-filtered when outputting to the tabular Varscan2 format, but not when to VCF (using --output-vcf). Strand-filtering is turned off (the default). I examined this across versions of Vascan2, and it's consistent, and is not related to the strand filter fix in 2.4.0.
VarScan.v2.3.7.out_tab.log:1 were removed by the strand filter
VarScan.v2.3.7.out_vcf.log:0 were removed by the strand filter
VarScan.v2.3.8.out_tab.log:1 were removed by the strand filter
VarScan.v2.3.8.out_vcf.log:0 were removed by the strand filter
VarScan.v2.3.9.out_tab.log:1 were removed by the strand filter
VarScan.v2.3.9.out_vcf.log:0 were removed by the strand filter
VarScan.v2.4.0.out_tab.log:1 were removed by the strand filter
VarScan.v2.4.0.out_vcf.log:0 were removed by the strand filter
VarScan.v2.4.1.out_tab.log:1 were removed by the strand filter
VarScan.v2.4.1.out_vcf.log:0 were removed by the strand filter
VarScan.v2.4.2.out_tab.log:1 were removed by the strand filter
VarScan.v2.4.2.out_vcf.log:0 were removed by the strand filter
VarScan.v2.4.3.out_tab.log:1 were removed by the strand filter
VarScan.v2.4.3.out_vcf.log:0 were removed by the strand filter
VarScan.v2.4.4.out_tab.log:1 were removed by the strand filter
VarScan.v2.4.4.out_vcf.log:0 were removed by the strand filter
The 2.4.0 release notes state:
This means that even though the default is to have the strand filter turned off (
--strand-filter 0), it may get applied in prior versions, with variants "strand filtered" erroneously. This was "fixed" in2.4.0. Nonetheless, we found that this "fix" gets rolled back in2.4.1and2.4.2, and then re-applied in2.4.3. Likely this fix got rolled back along with the changes with "the waynormal_reads2are counted". From the 2.4.0 release notes :Nonetheless, we found a variant that is strand-filtered when outputting to the tabular Varscan2 format, but not when to VCF (using
--output-vcf). Strand-filtering is turned off (the default). I examined this across versions of Vascan2, and it's consistent, and is not related to the strand filter fix in 2.4.0.