zhanglabpopgen.bsky.social
Assistant Professor at the University of Michigan. Popgen/Evolgen, admixture everything, ML etc. More on www.zhanglabpopgen.org
13 posts
207 followers
214 following
Regular Contributor
Conversation Starter
comment in response to
post
Thank you!
comment in response to
post
Thanks Bernard!!
comment in response to
post
n/n This was my last postdoc project that finally got completed with tremendous help from my postdoc mentors Kirk and Sriram, two UMich biostats MS students (Jiongxuan Yang and Lingxuan Zhu), and many colleagues and friends in MI and CA (esp. Jazlyn Mooney). Can't thank them all enough!
comment in response to
post
6/n Overall, we show that recessive deleterious mutations not only exist on the human genome, but also are not uniformly distributed and are mostly found in regions regulating metabolism and immune functions (our detected regions include the HLA cluster)
comment in response to
post
5/n We further investigated the biology of such regions and show they are 1) depleted of haploinsufficient genes, 2) depleted of long ROHs, 3) enriched with UKBB variants with evidence of fitting non-additive model, 4) experienced weaker historical BGS, and 5) more tolerant to LOF mutations
comment in response to
post
4/n we applied DominL to 7 non-African pops in 1KG, and show that approx. 3-9% of the human genome is enriched with recessive deleterious mutation! This is important because dominance is very understudied in humans due to many constraints with methodology (eg. can't distinguish h from hs compound)
comment in response to
post
3/n our method DominL shows decent power and robustness at finding genomic regions enriched with recessive deleterious muts (or close to fully recessive such as h<0.1; even if the model is trained using two extreme h values) at 1MB resolution, with exceptional accuracy in exon-dense regions
comment in response to
post
2/n we developed an ML approach using Neanderthal ancestry as one of the key features and trained 13 ML classifiers with simulations using fully recessive vs fully additive deleterious mutations. After a rigorous performance comparison and feature selection, we picked a winner ML model
comment in response to
post
1/n This is a spin-off of some earlier papers showing that recessive deleterious muts can confound adaptive introgression due to heterosis. And we ask a reverse-engineering Q: given the empirical archaic ancestry distribution, can we infer if any genomic region is enriched with recessive mutations?
comment in response to
post
Superb!!! Congratulations Fernando!!!!
comment in response to
post
Congratulations!! That's a fantastic department and they are lucky to have you!
comment in response to
post
Congratulations Vince!!