#ashg24 — Public Fediverse posts

PacificBiosciences/HiFi-SVTopo: Complex structural variant visualization for HiFi sequencing data github.com/PacificBiosc... ...being presented at #ashg24 this year. 🧬🖥️

Continued 2: GDS: Mendelian randomization design: Z → X → Y ← C Z: genetic variant C: confounder X: exposure of interest Y: outcome #ASHG24 🧪🧬🖥️

Not actually #ASHG24. Anyway, continuing on. #

GS: Mendelian randomization introduced because of the dramatic failure of other attempts to find causality from observational methods. #ASHG24

Now: @mendelrandom.bsky.social (GS): Mendelian randomization: what it was, what it is, and what it should become #ASHG24 🧪🧬🖥️

NEWS! Check our wrap-up about #ashg24 https://shorturl.at/WcTiz

Something is missing? Let us know!

🙌 Thanks, ASHG, for giving us the floor to share updates with #genetics community, delivered by our colleague @mariacerezo.bsky.social

#OpenAccess #gwas #gwasdatahero

Stronger together @ebi.embl.org @NIH

#ASHG24 There were two more talks in that session, but I hit my limit for the day. My flight hopefully leaves early tomorrow. That should be a wrap for me this year. See you next year for #ASHG25.

#ASHG24 AU: Resolved so far: 3q29 deletion, 16p11 duplication, 16p11 deletion, 1q21 duplciation, etc.

Example of 16p11 duplication. Again break points are on transposons.

#ASHG24 AU: Ignore the reference and assemble the duplication haplotypes directly.

Different 22q11.2 patients have different deletion-flanking regions.

CTLR-Seq assemblies let you find break points. Guided by pan-genome assembly.*breakpoints localize to transposon sequences.*

#ASHG24 AU: Segmental duplications create loci of genomic instability. Standard reference isn't useful in many of these regions [ unresolved. ]

Can't read through with short reads. or linked reads. or long reads. problematic region.

CRISPR-Catch method. in vitro cutting outside of segmental duplications. remove rest of genome from the region. Then long-read of up to 2Mbp fragment. Resolves segmental duplications and find break points.

#ASHG24 AU: large neuropsychiatric dx associated cnvs share characteristics.

Large. Can be complex with unknown break points.

Example 22q11. 3Mbp. ~50 genes affected. Up to 0.1% of live births. Pleiotropic with high risk of SCZ.

#ASHG24 AU: Alexander Urban. Revealing the exact breakpoints and sequence rearrangements of recurrent, large neuropsychiatric copy number variations (CNVs) at single base-pair resolution using CRISPR-targeted ultra-long read sequencing (CTLR-Seq).

psychiatric disorders have a strong eentic component. GWAS yielded about 300 loci for schizophrenia (SCZ).

[ nice plot showing MAF versus effect size. GWAS hits low effect, but very rare with large effects. ]

#ASHG24 ZT: Incorporate new mutations leads to tracking more recent ancestry.

Several signals across ancient groups. Excess recent ancestry between ancient and modern samples: TCHH, LCT, MCM5, CCR5, TLR1/6, MHC, etc.

#ASHG24 ZT: Focused on West Eurasian ancient genomes. Vikings. Romans. Etc.

Looked at Saxon migrations. Irish Medieval closest to modern Scots.

Look at group migration patterns. Zoom in on area of UK with lots of nordic place names. People there close to viking ancient DNA genomes.

#ASHG24 ZT: TheaDNA is computationally eficient. Linear for ancient samples. Sublinear for the modern walkthroughs.

Simulations to access accuracy. For segments over one cM almost always in right region of closet relative set.

Applied to ARGs for 487k from UKBB and 4800 ancient DNA samples.

Get pairwise measure of sharing between ancient and modern individual over a user set timespan.

#ASHG24 ZT: ThreaDNA method.

Input unphased ancient DNA. Trace two paths through phased modern panel. Output is amalgamation of those segments, i.e. a potential ancient DNA haplotype determined from most likely path through modern genome.

Infer age of segment. Estimate the age given the length. Longer segments should be younger.

#ASHG24 ZT: Ancestral Recombination Graph (ARG) are complex to calculate. Can use them to infer evolutionary history.

Existing methods are difficult to scale. Require high coverage genomes (not available for most ancient genomes). Require phasing (unreliable for ancient genomes).

New method called ThreaDNA.

#ASHG24 ZT: Zoi Tsangalidou. Inference of genome-wide genealogical relationships between ancient and modern individuals.

Genealogical relationship over time can be a tree. Two individuals at a locus eventually coalesce at a common ancestor.

How do you do it genome-wide?

Sequence of trees across the genome. Make an ancestral combination draft. Compact representation of genealogical history. Can incorporate mutation information.

#ASHG24 SM-W: PRS predicts a 25% increase in inpatient costs form middle to top decile in finngen.

evidence of indirect parental effect but prs score attenuation is modest.

#ASHG24 SM-W: rare coding variants for actionable conditions have large impact on inpatient healthcare costs. BRCA1, BRCA2, MSH2, APC.

Example BRCA2 had much higher costs than older individuals with BRCA2.

Adjust for participation bias in UKBB helps to generalize cost estimate for whole UK population. Cost of annual inpatient for BRCA2 LoF mutations among 40-69 yo could be 27M monies annually

#ASHG24 SM-W: Example - missense in SH2B3 associates with increased prescription drug costs. rs3184504. Involved in immune fxn and thyroid conditions. SNP associated with 60% increased usage of thyroid preparations.

Example - PITX2. rs6854111. Developmental transcription factor. Associated with increased inpatient costs.

#ASHG24 SM-W: GWAS for healthcare cost. 53 for inpatient costs. 187 for prescription drug cost.

Example HLA associated with inpatient costs [ autoimmune dx? ].

HLA also for prescription cost. High genetic correlation among studies despite substantially different median costs in different countries.

Nearly 1/3 of loci have age- *and* sex-specific effects. Example CHEK2. rs62237617. Higher in females (associated with breast cancer)

#ASHG24 SM-W: A small proportion of individuals contribute disproportionally to healthcare cost in the UK based on UKBB.

There is large variability in costs across healthcare systems and countries. 11 cohorts from 7 countries. Inpatient costs available for 1.3M individuals. Prescription drug costs for 1.2M.

#ASHG24 SM-W: Healthcare cost is directly calculated from health record using reimbursement algorithms.

2004-2019. Four major cost phenotypes. 1. Inpatient. 2. Inpatient + Outpatient. 3. Prescription drugs. 4. primary care appointments.

DRG codes to cost from healthcare prospective.

#ASHG24 SM-W: Sebastian May-Wilson. Quantifying the impact of genetic variation on healthcare cost across 1.5 million individuals from 13 studies and 8 countries: the GenCost consortium.

why do we care about the genetic basis of healthcare cost? longitudinal proxy of overall health. Increased costs a marker of frailty. Can help determine cost-effectiveness of genetically informed public health interventions.

#ASHG24 Featured plenary abstract session.

#ASHG24 that's a wrap for omics in complex disease.

#ASHG24 PA: More SVs in the ancestries you know to expect more variation (African ancestry > European).

Examples - 352 bp duplication, 4274bp deletion fused together.

NBPF8 complex rearrangement. Large inversions plus deletions in the same area.

Not all complex alignemtn patterns are complex SVs. Balanced inversion in a complex loci could be called as a complex rearrangment depending on the tool.

#ASHG24 PA: large variants fragment alignments. Trace reference and assembly over SVs. Anchor on alignments. Try all variant types per edge. Solve the graph (edge weights are the alignment score).

Apply to their 65 samples. Several hundred complex rearrangements. Most are near telomere and centromere. Drop to 81. About 68 if you exclude high level VNTR enrichments.

#ASHG24 PA: Peter Audano. Detecting large complex structural variants from human genome assemblies.

Based on long reads. HiFi and ONT. With long reads you can get SVs, small indels, and SNVs.

Part of the human genome structural variation consortium. Used phased assembly variant caller. 65 samples (130 phased haplotypes.

Most SVs are simple, but some are complex combinations of insertion, deletion, and inversion.