Publish Version III of the SNP+TR reference haplotype panel

[gymreklab]

The Version III files are the same as VII, with the following updates to facilitate use in downstream imputation pipelines:

1. Remove TRs for which the REF allele does not match the expected sequence based on CHR:POS
2. For each TR, remove alleles with 0 count (with the exception of the REF allele which should always be present even if it has count 0)
3. Remove TRs which have more than 100 alleles.
4. Remove TRs which have less than 2 alleles.
5. Remove the DS/GP fields which are large and not used by downstream steps.
6. Add unique IDs for each TR of the format EnsTR:CHROM:POS. For TRs with the same CHR:POS, add the duplicate number of the TR following format: EnsTR:CHROM:POS:Duplicate_num.
7. Add VT field, set to VT=TR for TRs and VT=OTHER for other variant types
8. Add the .bref format files which have the same information as the VCFs but can improve Beagle imputation performance.

The script scripts/fix-ref/fix_ensembletr_snpstr_reference.py makes these changes.

VII files (and V1 genotype files) have been moved to archive_ensembletr_datasets.md so the main README doesn't get too cluttered.

[gymreklab]

@heliziii you can review but let's hold off on merging until @yli091230 updates links to the new ref files

[yli091230]

@heliziii I just updated links and README file. It should be good to go.

[heliziii]

Are we publishing version 3 or 4?

[heliziii]

ah I saw that we archived version 3, can you remind me what changed from version 3 to 4?

[yli091230]

I not sure. The version 3 have some issues with missing reference alleles. @gymreklab , which version number should we use?

[yli091230]

In version 3, some REF alleles are missing, due to no REF allele detected. This will cause error in the downstream analysis. To fix it, we always keep the REF alleles in version 4.

[heliziii]

does this mean at least one alternative alleles?

[yli091230]

Yes.