Statgen specific tutorial

[hyanwong]

The last one of the subaudience tutorials to tick off is the one for statistical geneticists. This issue can collect some ideas.

I have been chatting to Arni Gunnarson, and he suggested that the following topics would be of interest to a statgen researcher looking into tskit (here biased towards the ARGneedle - type approach, obviously)

• fast GRM calculation (see tskit's `genetic_relatedness()` versus eGRM (Fan et al. 2022) tskit#2603 - this is not ready yet though)
• Luke O'Connor / Wilder's "LDGM" stuff, now in Nature Genetics(!): https://www.nature.com/articles/s41588-023-01487-8
• putting phenotypes onto individuals in a tree sequence (e.g. Daiki's tstrait)
• once phenotypes are associated with individuals in a tree sequence:
  • fast calculation of heritability
  • ARG-informed association analysis similar to that in the ARGneedle paper

[hyanwong]

Once we get Hanbin's tslmm published, this should give a great statgen angle. Also we should include running the Li & Stephens algorithm on a tree sequence, once that is documented (e.g. tskit-dev/tskit#452)

[hyanwong]

Also PCA would be good to include.

[hyanwong]

We should also say that tree sequences can be efficient converted to GRGs and Linear aaRGs, allowing the efficient statgen functions of those libraries to be used as alternatives to built-in Tskit 'site-mode' functions, which includes standard GWAS.

[hyanwong]

Also GWAS via https://tskit.dev/tskit/docs/stable/stats.html#trait-correlations

[gregorgorjanc]

We will be hiring on the BBSRC/NSF grant and that colleague could take a lead on this tutorial;)

[hyanwong]

That would be excellent. Thanks @gregorgorjanc