How to run joint variant calling in GATK?

Prepare Your Data

• Ensure your reference genome is properly indexed and prepared for analysis. You can use tools like `bwa` for alignment and `samtools faidx` for indexing the fasta file.

 

• Prepare a sample manifest file listing all the individual gVCF files that will be used in joint variant calling.

 

• Normalize the names and formats of BAM or gVCF files to ensure consistency and avoid potential issues during processing.

Generate gVCF Files

• Use GATK's `HaplotypeCaller` in GVCF mode to generate a gVCF file for each sample. This is accomplished with the `-ERC GVCF` option:

 

• Command: `gatk HaplotypeCaller -R reference.fa -I sample.bam -O sample.g.vcf -ERC GVCF`

Combine gVCF Files

• Use the `CombineGVCFs` tool if necessary, to merge gVCF files from the same group of samples or to partition them for parallel processing.

 

• Command: `gatk CombineGVCFs -R reference.fa -V gvcf1.g.vcf -V gvcf2.g.vcf -O combined.g.vcf`

Perform Joint Genotyping

• Run `GenotypeGVCFs` on your combined gVCF file(s) to perform joint genotyping across all samples. This will generate a multi-sample VCF file:

 

• Command: `gatk GenotypeGVCFs -R reference.fa -V combined.g.vcf -O raw_variants.vcf`

Filter Variants

• Use `VariantFiltration` to apply hard filters to the raw variants based on recommended thresholds or the specific needs of your study.

 

• Command: `gatk VariantFiltration -R reference.fa -V raw_variants.vcf -O filtered_variants.vcf --filter-expression "QD < 2.0" --filter-name "LowQD"`

Review and Validate Results

• Examine the filtered VCF file, making sure to validate the integrity and quality of the called variants. Consider additional quality checks or downstream analyses as required.

 

• Utilize visualization tools such as IGV for manual validation of variant calls.

Document and Archive

• Document your pipeline, including all parameters and software versions used, ensuring reproducibility and compliance with best practices.

 

• Archive all raw data, intermediate files, and result outputs appropriately.