tanyaphung/gatk4_notes

Notes on GATK4

GATK4_notes

Notes on GATK4. Here, I'm interested in calling variants and outputing all of the sites (not just the variants).

1. Call variants per sample per chromosome to generate the gVCF files

• Tool: HaplotypeCaller
• --emit-ref-confidence BP_RESOLUTION

gatk HaplotypeCaller \
  -R ref \
  -I bam \
  -L chrm \
  --emit-ref-confidence BP_RESOLUTION \
  --output output

2. Consolidate GVCFs

• Tools: GenomicsDBImport or CombineGVCFs.

• The GATK4 Best Practice Workflow for SNP and Indel calling uses GenomicsDBImport to merge GVCFs from multiple samples. GATK highly recommends using GenomicsDBImport for consolidating (over CombineGVCFs). However, I find that when using GenomicsDBImport to consolidate GVCFs, the flag --include-non-variant-sites does not work in the GenotypeGVCFs step. If you do not need to include the non-variant sites in your VCF, I find that GenomicsDBImport and GenotypeGVCFs work well.

• As mentioned above, I find that in order for the flag --include-non-variant-sites to work in the GenotypeGVCFs step, the input consolidated GVCF file should be consolidated using the tool CombineGVCFs (rather than GenomicsDBImport).

• Using CombineGVCFs:

gatk --java-options '-Xmx10g' CombineGVCFs \
-R ref \
--variant sample1.g.vcf.gz \
--variant sample2.g.vcf.gz \
--output all.g.vcf.gz

• Using GenomicsDBImport:

gatk --java-options -Xmx8g \
GenomicsDBImport \
--genomicsdb-workspace-path my_database 
-L chrm 
--sample-name-map cohort_sample_map
--tmp-dir=tmp_dir 
--reader-threads 5

• The file cohort_sample_map is a tab-separated file where the first column is the name of the sample and the second column is the path to the GVCF file for that sample. Note that this has to be tab-separated or else it will fail.

3. Joint Genotyping

• Tool: GenotypeGVCFs

gatk --java-options '-Xmx8g' GenotypeGVCFs \
-R ref \
-V all.g.vcf.gz \
--include-non-variant-sites \
--output allsites.vcf