How is the ancestral state stored in sgkit?

[hyanwong]

Hi - off the back of some questions about ancestral state polarisation triggered in tsinfer (tskit-dev/tsinfer#517), I wondered how sgkit intends to store ancestral state information, and how it might be possible to interface that to tools such as tsinfer or Peter Keightley's est-sfs software?

[hammer]

@hyanwong thanks for the inquiry! We don't currently store ancestral state information in sgkit. If you have a use case in mind, we'd love to learn more about it so that we might incorporate it into our library.

[hyanwong]

Thanks for the reply @hammer. I definitely have a use case in mind: it's important for our tree sequence inference pipeline (and also for things like the unfolded SFS). I'm sure @jeromekelleher has a better grasp of how sgkit will interface with tsinfer, but this is certainly one of the pieces that needs to be considered. At the moment, we simply require the ancestral state to be the first specified allele (i.e. index 0 in the samples x sites matrix). But it's helpful to be able to repolarise the ancestral state too (hence the reference to est-sfs which is a method for inferring the ancestral state).

Let me know how much more I need to explain, or maybe we should wait until @jeromekelleher is back in action.

[hammer]

I think that’s enough information for now! I filed https://github.com/pystatgen/sgkit/issues/585 with an idea for an approach that might work and will get @jeromekelleher’s thoughts when he returns. Thanks again for bringing this requirement to our attention.