README For Variant Calling by VarScan on TACC (Stampede2)

Written by : Taslima Haque Last Modified: 03/03/2023

Contact : [email protected]

Expected parametes to set:

Here is the example header of each script which expect the following variables: (Please set this with our own personal work directory on Stampede2)

refDir=/some/path/ #directory where the reference genome file will be
ref=/some/file # Name of reference genome file
outDir=/some/path #output directory. It must be created before running the script
met=/some/path/Pvirg_48_midwest_metadata_mod.csv
TMP=/some/path/Temp

Here is the sample of Meta file that is tab separated with fields of unique sample ID, sample name, and other meta info:

IELF,J020.A,Natural Collection,Grown from GRIN Seed Packet In Austin,\
Florida,27.197548,-80.252826,Unknown,Midwest,Midwest
IELJ,J036.A,Natural Collection,Grown from GRIN Seed Packet In Austin,\
South Dakota,44.520755,-99.200387,Upland,Midwest,Midwest
IELK,J037.A,Cultivar,Grown from GRIN Seed Packet In Austin,\
South Dakota,46.388289,-100.98355,Upland,Midwest,Midwest
IEMC,J254.A,Natural Collection,Supplied by Mike Cassler,\
Wisconsin,42.496,-87.808,Unknown,Midwest,Midwest

Tools requried:

  -- bwa-mem2 (install locally)
  -- samtools
  -- picard
  -- varscan (install locally)
  -- bcftools (install locally)

The pipe expect the output directory is the 1st level directory that already exists & have following directories:

ls  outDir/

          RAW
          MAP_SORTED
          MAP_SORTED_DEDUP
          VarScan
          VarScan_Filter
          MergedVCF

Each of these steps create a specific param file which you need to run by slurm. The slurm file is provide with the pipe named as "slurm.sh"

Step 1: Map and filter

Index the reference genome if needed

bwa-mem2 index Reference.fa

  sh 01-BWA2-Mapping-Filter-Sort.sh

This step will generate a param file named as "bwa2-sort.param". We are going to run this on "skx-normal" queue with 12 hours limit while each job will take one entire node. Using "skx-normal" instead on "normal" queue will be quick due to faster clock speed.The max it took me to run the largest sample was 8 hours therefore 12 hours should be a safe limit. The following command should work for 48 samples:

sbatch -t 12:00:00 -N 48 -n 48 --ntasks-per-node=1 -p skx-normal slurm.sh \
bwa2-sort.param

Step 2: Remove Duplicates

sh 02-Dedup.sh  

This step will generate a param file named as "dedup.param". We are going to run this on "normal" queue with 15 hours limit while each job will take one entire node. This step is limited mostly for memory than CPU therefore on "normal" queue it would be cheaper and not much less faster than "skx-normal" queue. We will keep the max time as the upper limit which is 48 hours. The following command should work for 48 samples:

sbatch -t 48:00:00 -N 48 -n 48 --ntasks-per-node=1 -p normal slurm.sh \
dedup.param

Step 3: Call variants

sh 03-Samtools-Varscan.sh

This step will generate a param file named as "varscan.param". We are going to run this on "long" queue with 120 hours limit. We will run 6 jobs on a single node. The max it took me to run the largest sample was 62 hours but we will set it to the max upper limit. The following command should work for 48 samples:

sbatch -t 120:00:00 -N 8 -n 48 --ntasks-per-node=6 -p long slurm.sh \
varscan.param

Step 4: Filter variants

sh 04-VCFFilter-Rename.sh

This step will generate a param file named as "filvcf.param". We are going to run this on "normal" queue with 9 hours limit. We will run 8 jobs on a single node. The max it took me to run the largest sample was 6 hours therefore 9 hours should be a safe limit. The following command should work for 48 samples:

sbatch -t 09:00:00 -N 6 -n 48 --ntasks-per-node=8 -p normal slurm.sh \
filvcf.param

Step 5: Merge variants

sh 05-MergeVCF.sh

This step will generate three param files named as "bcfindex1.param", "bcfmerge.param", and "bcfindex1.param". The computational requirements of this step entirely depends on the number of samples we are merging and may require optimization.