How to create phg with existing VCF file?

[yifangt]

I'm trying to get phgv2 running for Imputation, but have got lost with the guide because I already have a VCF (of 522 samples) generated with GATK4 pipeline. My questions are:

1. How to create the range file in BED format?
   The reference is not a model plant and does not have an annotation yet (so that there is not a .gff file to create the range file in BED format), I was thinking to use the VCF file with certain filter criteria such as MQ, DP, QUAL etc. to create a bed file for the ranges. Is this acceptable? Or, can I simply use the whole reference genome for the range file?

2. what should be done instead of "align assemblies" in order to create the hVCF and gVCF against the reference genome? I was thinking to use the existing re-calibrated BAM files from bwa alignment for the GATK4 SNPcalling. But, the phg gvcf2hvcf command seems simpler but confusing to me. Which should be used?

3. In order to load VCF data into DBs, phg create-ref-vcf and phg create-maf-vcf seems needed to be run first, right? Then this question gets into a loop of above 2 questions.

If my above thinking/logic is totally wrong, what is the correct protocol to run Imputation on existing VCF of multiple samples? Really appreciate the community can give me some advice to sort out the pipelines. Thanks a lot!

[zrm22]

Hello,

The PHG is designed to start with assemblies and build up from there. We currently do not have a way to go from a VCF file into PHG as assembly sequence is at its core.

For your questions:

1. If you do not have a annotated genome yet you can provide a BED file based on whatever slicing criteria you wish but we typically suggest that you choose breakpoints that are heavily conserved in your species.
2. You need to do the align-assemblies step. For the PHGv2 there is no way around this and have it work correctly. The PHGv1 was designed to create haplotypes based on WGS short reads but having a set of high quality assemblies tends to give much more reliable results. The PHGv1 is currently not under active development as well. PHGv2s gvcf2hvcf is designed to convert between the two sets of files if generated through the align-assemblies anchor wave pipeline.
3. And yes create-ref-vcf and create-maf-vcf need to be run for v2 and create-maf-vcf specifically is what takes the anchor wave .maf output and converts it into both a gvcf and an hvcf.

I think basically the PHGv2 is not designed to handle your use case.

[yifangt]

Thank you very much for this clarification. I really appreciate your instant response.