VAF ~ 1%

[lmanchon]

--Hi,

how to set Vardict to detect variants with VAF ~ 1%, which parameters can increase sensitivity ?

than you --

[PolinaBevad]

Hello! Usually you can set -f 0.01 and this will work as originally VarDict was developed to call low AF variants (<0,1%). Most of the default quality options, mean position, realignment module, etc. should work as is with Illumina sequencing data.

[lmanchon]

--Hi, thank you for your answer. and i have just another question: is there a somatic and germline search mode like with Strelka software ? thank you --

[PolinaBevad]

Default 'single sample' mode is when you provide one BAM file. It will call all the variants without any germline or somatic assumptions. Also there is 'paired' mode when you provide both normal and tumor BAM files so you can classify variants as somatic/germline/LOH/etc: https://github.com/AstraZeneca-NGS/VarDict/wiki/Variant-classification-in-paired-analysis

[lmanchon]

can you suggest me parameters to increase the sensitivity of detection of variants quite to remove filters on the quality of sequences or alignments ?

[PolinaBevad]

Default base quality threshold is 22.5 (-q option), and also ratio of high-quality reads to low-quality reads (qratio) is 1.5 (-o option) so you can adjust it if you know that quality is underestimated by sequencer. We have the explanation of default filtering algorithm in Java version docs: https://github.com/AstraZeneca-NGS/VarDictJava/blob/master/Readme.md#variant-filtering Also there is a useful pileup (-p option) mode that disable all filters so you can see variants values to check why they were filtered out. It will also output reference bases but this can be filtered out downstream.

[lmanchon]

ok thank you so much for this help

[lmanchon]

typicaly VarDict use recalibrated bam as input or not, i know freebayes prefer not recalibrated bam