iiinova — where your digital self keeps evolving

Wolfram Höps @wolfram-hops.bsky.social · 06/03/2026

(17/17) Please note this "skeetorial" is (a) only my personal take and (b) not optimized for scientific accuracy. Please consult the paper for hard&dry facts :) Looking forward to any

110

Wolfram Höps @wolfram-hops.bsky.social · 06/03/2026

(16/17)(3) Risk is predictable and varies per individual/population (4) Human evolution really likes the segmental duplication "web" and refuses to get rid of it, despite creating instability. (At least in the reference-like state, repeats are in inverse orientation so not much risk.)

100

Wolfram Höps @wolfram-hops.bsky.social · 06/03/2026

(15/17) Some takeaways: (1) 15q13.3 CNVs arise from alleles that are 'primed' for NAHR, especially via inversions. (2) Instances of the 'same' CNV can vary (e.g. only some affecting CHRFAM7A, sth i did not talk about here enough). Consider them 'pseudo-recurrent'.

100

Wolfram Höps @wolfram-hops.bsky.social · 06/03/2026

(14/17) If these repeats lead to CNVs, why did evolution not get rid of them? DongAhn Yoo shows that these duplications are human-specific and evolved recently in human evolution. There seems to be a selective advantage to them that outweighs the CNV risk. Many 'brain-genes' in the region...

100

Wolfram Höps @wolfram-hops.bsky.social · 06/03/2026

(13/17) As a side note: at least the 'large' inversion can be pretty well imputed from SNPs. So one can test/refine many of our claims even in short read data. Here too, more datasets will help, and we're looking into that.

100

Wolfram Höps @wolfram-hops.bsky.social · 06/03/2026

(12/17) Given what we've learned about CNV formation, we can classify alleles with respect to how likely they are to lead to CNVs. Turns out: The risks differ by population. e.g., Europeans are >10 times more predisposed than East Asians. We are searching for clinical incidence data to confirm this.

100

Wolfram Höps @wolfram-hops.bsky.social · 06/03/2026

(11/17) How does 15q13.3 look like in healthy individuals? We looked at 581 assemblies from @humanpangenome.bsky.social and HGSVC. The short answer: haplotypes are inverted, deleted, duplicated left right and center. We count at *least* 18 distinct configurations.

100

Wolfram Höps @wolfram-hops.bsky.social · 06/03/2026

(10/17) We can formalize this a bit more: There are 2 different large inversions, and 2 different possible breakpoints; leading us to predict 8 'flavors' of CHRNA7 CNVs - three observed in our data. This strange rearrangement was proposed before (Shinawi 2008), but now we can see it in seq data.

100

Wolfram Höps @wolfram-hops.bsky.social · 06/03/2026

(9/17) What about the smaller 'CHRNA7' CNVs? Here things are even wilder. This is not a simple deletion - rather, there is ectopic recombination *inside* a locus-spanning heterozygous inversion - creating a weird fusion that simultaneously deletes sequence (CHRNA7) and duplicates others (ARHGAP11B).

110

Wolfram Höps @wolfram-hops.bsky.social · 06/03/2026

(8/17) In three cases, the breakpoints point to a 2 kbp recurrent recombination site with a strong PRDM9 motif. Two cases dip into 'GOLGA8' motifs which are known to promote instability. Biology being biology, one CNV falls completely out of the norm, with a cpx, non-NAHR event.

110

Wolfram Höps @wolfram-hops.bsky.social · 06/03/2026

(7/17) BP4-BP5. For de novos, we can compare parent alleles (pre-CNV) with child alleles (CNV) to find breakpoints. Turns out: in 5/6 cases, an inversion brings the red repeat pair ("SDB") in same orientation-> substrate for NAHR->CNV. On a 'reference' haplotype, this particular CNV is impossible.

100

Wolfram Höps @wolfram-hops.bsky.social · 06/03/2026

(6/17) We recruited ten patient-parent trios with 15q13.3 CNVs to find out, sequenced with @pacbio.bsky.social + @nanoporetech.com and generated de novo assemblies for 6 'BP4-BP5', 4 'CHRNA7' CNVs. Which sequences (genes!) are affected? Are the 'same' CNVs always identical? Breakpoints? Mechanisms?

100

Wolfram Höps @wolfram-hops.bsky.social · 06/03/2026

(5/17) NAHR can explain CNVs - but only in the presence of long repeat pairs in the same direction. in 15q13.3, all major pairs are inverted. So why still CNVs? What's more, CHRNA7 CNVs are not at all between same repeat pairs. Why CNV?

100

Wolfram Höps @wolfram-hops.bsky.social · 06/03/2026

(4/17) Like many recurrent CNV loci, 15q13.3 houses a network of long repeats -segmental duplications- that (a) predispose to CNV formation via non-allelic recombination (NAHR) and (b) make sequencing a nightmare because (short) reads won't map uniquely. Result: CNVs+breakpoints remain "blackboxes"

100

Wolfram Höps @wolfram-hops.bsky.social · 06/03/2026

(3/17) 15q13.3 is among of the most unstable loci in humans; with an incidence of e.g. ~1:5000 live births for the larger "BP4-BP5" deletion/duplication alone. Carriers are very differently affected - some heavily, others not at all.

100

Wolfram Höps @wolfram-hops.bsky.social · 06/03/2026

(2/17) Shout out to a great team including David Porubsky, DongAhn Yoo, @michelledegroot.bsky.social @cpschaaf.bsky.social @nicbrunetti.bsky.social @humanpangenome.bsky.social Evan Eichler, Christian Gilissen, and more unfortunately not on this platform. Esp.: patients+families who participated.

200

Wolfram Höps @wolfram-hops.bsky.social · 06/03/2026

🔄 Inversions predispose to recurrent deletions and duplications in chromsome 15q13.3. 🔁 Using de-novo assemblies of 10 patient-parent trios, we investigated how recurrent copy-number variants (CNVs) in the 15q13.3 locus arise. www.biorxiv.org/content/10.6... A brief tour (1/17)

1104