Tumor-only mutation calling pipeline

Workflow for SNV calling on tumor-only targeted NGS data

Currently, there are 2 preprocessing steps on the bam files (but eventually these could be moved to an 'post-alignment' processing workflow)

• target region filter
• LoFreq indelquality calculation

Includes different steps:

1. LoFreq variant calling

• followed by 'lofreq filter' based on NGS quality metrics, including: minimal coverage --min-cov 20
  cap coverage at depth --max-depth 1000
  minimal mapping quality --min-mq 30
  minimal base quality --min-bq (30 / 20)
  minimal alt base quality --min-alt-bq (30 / 20)
  significance --sig 0.01
  minimal VAF --af-min 0.05 \ strand bias p-value --sb-alpha 0.05 \ strand bias for indels --sb-incl-indels \

• TODO: Homopolymer filter for Lofreq: 'HRUN= ' in vcf

--snvqual-thresh 77 --indelqual-thresh 61

2. Apply LoFreq Panel Of Normal (PON) Blacklist

• to remove all artificial/germline calls from the PON dataset (todo: determine what cutoff to use: filter out all calls present in at least 2 PON-cases)

3. Variant Annotation
4. Variant Discrimination

• distinguish snp/somatic
• optionally: from all somatic calls: distinguish driver/passenger (using MutSig2CV)
• distinguish functional/non-functional somatic
• optionally: from all functional somatic SNVs, select only the significant driver genes, based on list of significant driver genes from MutSig2CV-output