The genomic landscape of Ménière's disease: a path to endolymphatic hydrops.

Data

Data: s3://menieres-analysis-results/joint_genotype/2022-04-30_friedman_joint_metadata_04182022/ Final data on CCBB S3: s3://ccbb-projects/2021/20211026_friedman_menieres_disease_wgdnaseq/final_variants_08242022/

This repository should demonstrate how to transform raw WGS fastq files into actionable, rare, deleterious variants in the context of Meniere's disease.

Steps

1. Call variants on each sample using bcbio, implementing GATK's best practices.
2. Fix header of VCF files for joint genotyping (possibly a bcbio fault).
3. Joint genotype all samples.
4. Variant Quality Score Recalibration (VQSR).
5. Normalize variants - store variants at one record per line in VCF.
6. Annotate with VEP.
7. Convert VEP VCF to individual sample MAF.
8. Filter variants. for maf in $(ls */*/*maf); do out=$(echo $maf | sed 's/vep/vep.coding/'); awk -F '\t' '{ if($1 != "Unknown") { print }}' $maf > $out; done
9. Annotate with gnomAD GENOME.
10. Further filter for rare and deleterious variants.

1. Variant Calling

Input: *fastq.gz

Output: *-joint-gatk-haplotype-annotated.vcf.gz

You will first have to create a list of samples to run, and then execute run_germline.sh. This script submits germline.sh for each sample.

#!/bin/bash
batch=$1
s3Download=s3://menieres-raw-data/batch_$batch
s3Upload=s3://menieres-analysis-results
complete=$(aws s3 ls $s3Upload | grep PRE | awk '{print $NF}')
inprogress=$(qstat | grep ubuntu | awk '{print $3}')

for sample in $(cat /shared/workspace/projects/friedman/metadata/friedman_wgs_samples_batch$batch.txt); do
    if ! echo $complete $inprogress | grep -w -q $sample; then
        metadata=/shared/workspace/projects/friedman/metadata/20220125_friedman_wgs_"$sample".csv
        echo -e "samplename,description" > $metadata
        echo "/scratch/germline/$sample/"$sample"_R1.fastq.gz,$sample" >> $metadata
        echo "/scratch/germline/$sample/"$sample"_R2.fastq.gz,$sample" >> $metadata
        qsub \
        -N "$sample" \
        -v sample=$sample \
        -v s3Download=$s3Download \
        -v s3Upload=$s3Upload/$sample \
        -v metadata=$metadata \
        /shared/workspace/projects/friedman/scripts/germline.sh
    else
        echo "Sample $sample is complete. Skipping."
    fi
done

2. Fix VCF header

Input: *-joint-gatk-haplotype-annotated.vcf.gz

Output: *-joint-gatk-haplotype-annotated-fixed.vcf.gz

For some reason, bcbio creates both INFO/DP and FORMAT/DP in the VCF header. The joint genotyping code doesn't like this, so we have code to fix it. Execute run_remove_info_dp.sh, which runs remove_info_dp.sh for each sample and looks like this:

#!/bin/bash
#$ -cwd
#$ -pe smp 4

export PATH=/shared/workspace/software/bcbio/anaconda/bin:$PATH

workspace=/scratch/$sample
mkdir -p $workspace

echo "downloading "$sample"-joint-gatk-haplotype-annotated.vcf.gz"
aws s3 cp s3://menieres-analysis-results/$sample/ $workspace --recursive --exclude "*" --include "*gatk-haplotype-annotated.vcf.gz*"
mv $workspace/*/*gatk-haplotype-annotated.vcf.gz* $workspace

echo "removing INFO/DP"
bcftools annotate -x ^INFO/DP $workspace/"$sample"-joint-gatk-haplotype-annotated.vcf.gz -o $workspace/tmp.vcf.gz -Oz --threads 4
mv $workspace/tmp.vcf.gz $workspace/"$sample"-joint-gatk-haplotype-annotated-fixed.vcf.gz
echo "indexing $workspace/"$sample"-joint-gatk-haplotype-annotated-fixed.vcf.gz"
tabix -p vcf $workspace/"$sample"-joint-gatk-haplotype-annotated-fixed.vcf.gz

aws s3 cp $workspace/"$sample"-joint-gatk-haplotype-annotated-fixed.vcf.gz s3://menieres-analysis-results/$sample/$sample/
aws s3 cp $workspace/"$sample"-joint-gatk-haplotype-annotated-fixed.vcf.gz.tbi s3://menieres-analysis-results/$sample/$sample/

All it does is remove the INFO/DP part of the VCF header.

3. Joint Genotyping

Input: *-joint-gatk-haplotype-annotated-fixed.vcf.gz

Output: menieres-joint-gatk-haplotype-joint-annotated.vcf.gz

Make a metadata file named friedman_joint_metadata_04182022.csv containing the path of the fixed VCF, sample name, and batch name ("menieres-joint"). There should be one row per sample, formatted as follows:

samplename,description,batch
/scratch/joint/RF_0001-joint-gatk-haplotype-annotated-fixed.vcf.gz,RF_0001,menieres-joint
/scratch/joint/RF_0002-joint-gatk-haplotype-annotated-fixed.vcf.gz,RF_0002,menieres-joint

Add the path of the metadata file to run_joint.sh and execute. It will run bcbio's joint genotyping pipeline.

#!/bin/bash

metadata=/shared/workspace/projects/friedman/metadata/friedman_joint_metadata_04182022.csv
workspace=/scratch/joint

qsub \
-v workspace=$workspace \
-v metadata=$metadata \
/shared/workspace/projects/friedman/scripts/joint_genotyping.sh

4. VQSR

Input: menieres-joint-gatk-haplotype-joint-annotated.vcf.gz

Output: menieres-joint-gatk-haplotype-joint-annotated-vqsr.vcf.gz

We have to run Variant Quality Score Recalibration after joint genotyping. It may be a parameter you can add in bcbio for joint genotyping, but that was realized after the fact, so we are running it manually.

5. Normalize Variants and 6. Annotate with VEP

Input: menieres-joint-gatk-haplotype-joint-annotated-vqsr.vcf.gz

Output: menieres-joint-gatk-haplotype-joint-annotated-vqsr-normed.vcf.gz

The purpose of this is to normalize the VCF, so only one variant allele is represented per line so that when the variants are annotated, each gets their own true annotation because VEP only assigns one annotation per line. Execute norm_vep.sh to normalize and annotate with VEP.

chr1    15484   .       G       A,T

becomes

chr1    15484   .       G       A
chr1    15484   .       G       T

Output: menieres.norm.vep.vcf.gz

7. Convert VEP VCF to individual sample MAF.

Run vcf2maf.sh for each sample.

Input: menieres.norm.vep.vcf.gz

Output: a. $sample/vcf/$sample.norm.vep.vcf b. $sample/maf/$sample.norm.vep.maf

8. Filter Variants

Input: *.norm.vep.maf

Intermediate: *.norm.vep.coding.maf

Output: menieres.531.intermediate.tsv

Copy MAF from S3 to local directory and run:

for maf in $(ls */*/*maf)
    do out=$(echo $maf | sed 's/vep/vep.coding/'); awk -F '\t' '{ if($1 != "Unknown") { print }}' $maf > $out
done

RScript filterMaf.R

9. Annotate with gnomAD GENOME

Input: menieres.531.intermediate.tsv

Output: a. *.gnomadv3.vep.vcf b. *.gnomadv3.vep.maf

We have to annotate the variants with gnomAD Genome because VEP only annotates with gnomAD Exome. We need the WGS allele frequencies because there are huge discrepencies. To do this we have to:

1. Convert MAF back to VCF so we can annotate.

2. Annotate using bcftools to query gnomAD's remote DB. This takes a while so thats why we are doing it on filtered variants.

3. Concatenate and sort variants back together.

4. Reannotate with VEP.

This is done in maf2vcf_gnomadv3.sh

10. Further filter for rare and deleterious variants.

Input: menieres.531.gnomadv3.tsv

Output: menieres.531.gnomadv3.filtered.tsv

Filter for final set and visualize. This is done in RStudio with maftools in menierees_variant_summary.R

NOTE: You can see this isn't fully automated and there are some manual linker steps required.