hringbauer.bsky.social
Population Geneticist | Ancient DNA
Research Group Leader at MPI-EVA Leipzig
www.hringbauer.com
31 posts
612 followers
421 following
Prolific Poster
Conversation Starter
comment in response to
post
ARG-inspired. 😀 Interesting that they claim full ARG inference is not too useful for compression on slightly noisy genomic data.
comment in response to
post
In other contexts, such as in the search for rare, deleterious variants, an emphasis on ancestry groups is unwarranted. In particular, while rare nucleotide variants are often described as “ancestry-specific”, they will in fact be found in large enough samples of any ancestry.
comment in response to
post
So what happened to ARG-based compression? 🤔I think that was even a main reason to develop scaleable ARG inference (tsinfer)
comment in response to
post
Happy to join your meeting here at MPI-EVA Leipzig to discuss this work (if my schedule allows)!
comment in response to
post
Trying to understand the paradox:
Say you have a high variance and low variance measurement of the same unknown quantity of interest (exact variances unknown; both unbiased). You wouldn't take their average but just pick the low variance estimate- and omit data, right?
comment in response to
post
IBS segment sharing is not about nearest neighbors in a sample though - so not inherently biased for admixture modelling ;)
comment in response to
post
Link to the published paper:
doi.org/10.1093/gbe/...
comment in response to
post
Almost all common genetic variants are not population specific in humans. They only differ in allele frequencies!
We mainly used statistics summarizing many such small differences (f-statistics measuring shared drift) for the inference of those 45ky Eutopeans not contributing to later waves.
comment in response to
post
Ranis and Zlaty Kun still share the out-of-Africa bottleneck with modern groups! So while not directly ancestral, most coalescing is recent enough (<<100ky) for imputing common variants.
Similar to how one can impute modern East Asian genomes with European ref panels - despite the split 10kys ago.
comment in response to
post
Fascinating technical aspect: Imputing & ancIBDing 45ky-old human genomes works well. With modern ref panels and down to 0.1x WGS coverage! We used downsampling of the two high coverage genomes to prove that. 🎯 That direct experiment showcases the application range of those tool. ✅️✅️ (3/3)
comment in response to
post
Generally all Ranis individuals share very high background relatedness in form of long IBD (within Ranis and equally to ZK): We modelled an effective pop size of only Ne~200. These early European modern humans were small and mobile Europea hunter gatherer populations! (2/3)
comment in response to
post
Also curious about the effect of more & more LLN output on language. But my guess would be that certain phrases become en vogue - eg. everyone "breaks new ground" instead of the good old "we generated new insights" 😅
comment in response to
post
🤚I would be happy to be added.
Nice job!
comment in response to
post
It is large mass Graves embedded in a large urban Medieval cemetery - Black Death victims mostly lined up before decomposition. While there are no burial goods, we will make a detailed map of each individuals position (see Black shades in the map - those are the skeletons! ⬇️ )