Do freeBayes and platypus support gVCF (ie incremental variant calling for large sets of genomes)

Question

[William]

Do freeBayes and platypus support gVCF?

https://sites.google.com/site/gvcftools/home/about-gvcf

With GATK you used to have the 1000 + 1 genome variant calling problem, where you had to redo the whole expensive multi-sample variant calling (gatk haplotype caller) if you wanted to add a single genome.

Now GATK Haplotype caller can create intermediate gVCF files for each sample and do a much less expensive merging of the gVCF files to a multi-sample vcf.

https://www.broadinstitute.org/gatk/gatkdocs/org_broadinstitute_gatk_tools_walkers_variantutils_GenotypeGVCFs.php

Do freeBayes and and platypus also support this or are they planning to support this?

Answer 1

[donfreed]

My understanding is that freeBayes, platypus and samtools do not yet support gVCF files. However, you can work around this by calling variants from individuals files, merging the variant calls and then recalling the individual bam files at all sites called variant in at least one individual. See this excellent blog post

[William]

Well better than complete multi-sample variant calling but still very expensive I guess because of the centralized IO needed for recalling all variant sites in all the BAM files at the same time.

Answer 2

[William]

Basic support for gVCF has been added to Freebayes just this weekend. More support for gVCF will be added in the future.....

There are some inherent problems with gVCF and complex multi-allelic variant sites. For just snps it should work:

See also this issue at the Freebayes github site, the last comment is the update from Erik stating the (upcoming) gVCF support.

https://github.com/ekg/freebayes/issues/76