Feature Request: reading variants as sparse in pgenlibr

[joellembatchou]

Hi,

We rely on the C++ pgenlibr to read PGEN format genotype data in REGENIE. The current functions available (RPgenReader::ReadIntHardcalls/RPgenReader::Read) reads in the genotype data for all samples. In the case of rare variants, it seems PGEN stores the data sparsely based on the format documentation (i.e. only indices and genotypes of carriers are stored). Could this functionality be provided in the C++ pgenlibr, i.e. a flag identifying whether a variant is stored sparsely (given its index) as well as a function that returns the indices & genotypes (or dosages) of the carriers only?

Thanks,
Joelle

[chrchang]

Okay, I plan to provide this functionality on GitHub later this week, though the next CRAN release will wait till mid-year.

[chrchang]

HasSparseHardcalls(pgen, variant_num) now returns whether variant_num has a sparse representation. If it does, ReadSparseHardcalls(pgen, variant_num) returns an object where "sample_nums" has the sample indexes, and "counts" has the counts.

[joellembatchou]

Thanks that sounds great, looking forward to testing it!

[joellembatchou]

Small follow-up: is ReadMaybeSparseHardcalls() meant to be used if dosages are present? And if so, is there a function similar to HasSparseHardcalls() for dosages?

[chrchang]

ReadMaybeSparseHardcalls() ignores dosages if they're present. I will think about an appropriate way to provide this functionality for dosages.