Overall, the method is a scalable and interpretable way to look for haplotype homozygosity in biobanks. I have spent a lot of time implementing it to make it user friendly. Running this can be important to check for confounding: see our other preprint.
pubmed.ncbi.nlm.nih.gov/40672175/

While we have improved (made a more conservative threshold for the African American analysis), there are still unmodeled complexities and likely fine-scale structure that violates our assumption. Buyer beware if signal is weak (barely GW significant)!

We filter to have small admixture proportions in our African American analysis. We think that admixture is more likely to deflate our "selection signal", not inflate it. This is a robustness check.

We think that low-mappability, problematic regions are more likely to deflate our "selection" signal, not inflate it. See Appendix B. This is a robustness check.

There is a large signal of unusual haplotype structure at a chr16 locus in all groups. I also see this when analyzing other consortia data. We describe it in more detail as a complicated region enriched with deletions. Gusev et al have also reported on it.
academic.oup.com/mbe/article/...

This is a new conceptual figure which explains how to make our proofs. I hope this idea is helpful to other population geneticists.

This project completes my development of our suite of methods called isweep, which can detect recent selection and case-control associations, can narrow in on the selection signal, and has fast local simulations to perform statistical inference.

You can also run a fast scan with randomized phenotypes to double check if selection is confounding.

Sadly, the test is prone to confounding due to very strong positive and recent selection (see LCT in European ancestry cohorts). You can automatically check for this as a part of the automated workflow.

Two of the six genome-wide significant signals are about current therapeutic targets, e.g., at the NBAS gene.

We ran the scalable scan for three ancestry cohorts (European, African, and Amish) in the Alzheimer's Disease Sequencing Project.

The null model is again motivated by the Temple and Thompson (2025) central limit theorems, which is now published in Theoretical Population Biology.

The test is a two-sample modification of the one-sample selection scan from my prior work (Temple and Browning, 2025): high IBD rates differences in cases versus controls could indicate a genetic association.

And the tskibd paper itself on positive selection in Pf. is very interesting.
www.nature.com/articles/s41...

We discuss the use cases of local versus genome-wide IBD simulation. I'll say I have had a great time using msprime + tskibd for genome-wide simulations. We find it is accurate w.r.t. to our work. I highly recommend it.
github.com/bguo068/tskibd

One fantastic suggestion from both peer reviewers was to validate the accuracy of IBD segment length distributions w.r.t. existing tools. These studies are now included as Figures S7-10.

Correction: 3) should be approx. has the first-order Markov property. This is what I get for tweeting fast.