Genomic Variant Prioritization

Snakemake workflow post-genotype calling to prioritize disease-causing variants on biowulf2.

Quick-ish Test Start Using Demo vcf

• Log into your biowulf2 account.
• sinteractive
• mkdir -p ~/R/3.5/library
• module load R/3.5.2
• R
• devtools::install_github('davemcg/see_gem', build_vignettes=T)
• # THE ABOVE MUST INSTALL WITHOUT ERROR. IF IT DOES FIGURE IT OUT / ASK ME TO HELP.
• q()
• cd ~/
• mkdir -p ~/git
• cd ~/git
• git clone https://github.com/davemcg/variant_prioritization.git
• cd variant_prioritization
• sed -i 's/mcgaugheyd|guanb/YOUR_BIOWULF2_USERNAME/g' src/vcfanno_v4.conf
• cd variant_prioritization/tests
• sbatch --time=12:0:0 ../Snakemake.wrapper.sh config_variant_prioritization.yaml

Input

• VCF from NGS_genotype_calling Has to be bgzipped.
• PED with samples in VCF. The samples in PED and VCF must match. PED file has to be "\t" delimited. If header in PED, it has to start with #.
• SampleID in fastq files and PED files CANNOT contain "-" or "_". Can consider adding these characters in next version.
• "Default" Gemini quieries for samples and families will be included. Use

Set up

Copy src/config_variant_prioritization.yaml to your local folder and edit the ped field to give a path to your ped file. You will also need to edit the family_name to instruct Snakemake which families (must match ped family field, column 1) to create reports from. You can either give one family like so:

####- family_name: 'gupta_fam' - if you leave this blank (family_name: '') then only the GEMINI database will be created (no family reports) Or a list of families to process like so:- family_name: ['gupta_fam', 'smith_fam', 'chan_fam']

family_name will be generated from PED file by the pipeline

Install SeeGEM in R on biowulf2 to produce the html report.

• sinteractive
• module load R
• R
• devtools::install_github('davemcg/see_gem', build_vignettes=T)

Finally edit the first line of src/config_variant_prioritization.yaml to put your vcf (bgzip'ed and tabix'ed) in.

Run (in biowulf2)

freen to pick gpu p100 (default withouting specifying $2 below), v100 (need to edit cluster.json file), or k80 ($2 below). Currently using 4 gpu thus one node. When gpu node is busy, spliceai could take time to be started. sbatch --time=12:00:00 ~/git/variant_prioritization/Snakemake.wrapper.sh COPIED_OVER_YAML_FILE.yaml [optional: ~/git/variant_prioritization/src/k80cluster.json]

Visualization