Post processing of Somatic Variation Models

Contents

• Overview
  • Options that may be useful to you:
  • Less commonly used options:
  • Output

Overview

Post-processing of somatic-variation models can be accomplished through the use of the following Genome Model Tool: 'gmt somatic process-somatic-variation'

This tool deduplicates variant calls, filters out off-target sites, tiers the variants, adds dbSNP and GMAF information, etc. Appropriate files for manual review (XML and bed files) can also be generated.

If processing a large number of samples, you can give them all the same output-directory. Each model's output will be stored in a subdirectory with the sample name, and the manual-review files will all be grouped together in a review/ subdirectory.

This tool has many options, but a typical command might look like this:

$ gmt somatic process-somatic-variation --somatic-variation-model-id 12345678 --output-dir somatic-validation --add-dbsnp-and-gmaf --add-tiers --restrict-to-target-regions --create-review-files --tiers-to-review 1 --igv-reference-name=b37

Options that may be useful to you:

• --add-dbsnp-and-gmaf appends columns with dbSNP ids and global minor allele frequency (GMAF)

• --add-tiers appends a column containing the tier of each variant
  

• --create-review-files generates bed and xml files necessary to do manual review
  

• --tiers-to-review choose which tiers of variants should be placed into the bed files for review (default: 1)
  

• --igv-reference-name provide the reference name for the IGV session (most commonly: b37)
  

• --get-readcounts append readcounts from the normal and tumor bams for each variant
  

• --restrict-to-target-regions only keep calls in target-regions (as specified by the target_regions on the build)
  

Less commonly used options:

• --filter-regions

• --filter-sites Pass in either a list of variants to remove or a list of regions from which all variant calls will be removed.
  

• --sites-to-pass Pass in a list of sites that should be output, even if they would otherwise be filtered
  

• --required-snv-callers If set to a value greater than 1, requires that N variant callers independently call a variant before it is reported. Occassionally useful for filtering noisy data down to a manageable list.
  

• --sample-name This name will be used for the output directory, instead of the subject_name from the model.
  

Output:

• samplename/
  

•     snvs.indels.annotation Final output, containing filtered and annotated variants. Other columns are appended based on options above (tier, dbsnp, readcounts, etc)
  

•     snvs.indels.annotation.xls The same data as above, converted to xls format
  

•     snvs/  intermediate files from filtering, annotation, etc.
  

•     indels/ intermediate files from filtering, annotation, etc.
  

• review/ - files generated for manual review
  

•     samplename.xml
  

•     samplename.bed