American Society of Human Genetics meeting 2025.
Posted on :: Tags: conference, genetics, thoughts, spatial omics, idea flow

Table of Contents

TL;DR: Biased memos from ASHG 2025 talks/posters. AI/ML and new omics technologies always create new interests in the genetics community, but certain aspects of core genetic questions aren't easily changed. Large-scale biobank GWAS data became the norm; then what's next? Can we think of anything beyond polygenic risk predictions? Contextual information matters, but how so?

Oct 15

Non-linear GWAS methods

  • Refine additive association models with dominant and recessive models
  • Concerns over Type I error calibration
  • How is it different from model checking?

Context-specific polygenic score models

  • Fonseca, Andrew Dahl, U of Chicago
  • Improve prediction performance
    • Can contexts be anything?
  • Straightforward idea: calculate interaction GWAS
  • Build PGS models base off interaction coefficients
  • Can we come up with summary-based extensions?

How can include gene-gene interactions in MR?

  • No satisfactory answers

Oct 16

  • Squeezing AI models into genetic

Distinguishing causal variants underlying natural selection

  • Arya Rao, Sabeti Lab (Harvard), Steven Riley (Yale), DeepSweep

  • Selective sweep: SLC45A2, causal adaptive variants

  • Distinguish causal and neural-hitchhiker variants

  • Simulated data (coalescence model). Summary statistics.

  • Transformer-based prediction (do we need label data?)

  • GWAS and eQTL enrichment

  • But where is causal guarantee?

  • Question from audience:

    • Have you compared with colocalized or finemapping results?
    • Do ages of variants matter?

  • Takeaway: Probably useful... run simulation and train black box models.

PubMind language model trained on PUBMED

  • Train LLM on millions of abstracts
  • Existing resources: ClinVar, HGMD, LitVar2, ...
  • inference module: SNV, SV, CNV, Gene fusion
  • Focus on genetic evidence of pathogenecity
  • What are the queries?
  • LLaMa3.3, DeepSeek
  • Standardize disease name
  • https://pubmind.wglab.org
  • This is still a database... but the benefit of LLM is interactive API
  • We could find more, but how should we use the method?

Spatial 3D reconstruction of H&E data

  • Hu, Yu, Emory
  • Use long range dependency between snapshots
  • Multi-resolution alignment
  • Metamorphism (gradient flow)
  • Cool data: CyCIF + H&E on colorectal cancer
  • Cool 3D image... What did we learn?
    • Can we not find the same cell type identity mapping?
  • A question from moderator: Any downstream analysis?

Xenotransplantation

  • Eloi Schmauch, Keating, NYU
  • Pig kidney transplantation: pig vs. human cells?
  • Too much technology details? Clustering?
  • Visualization by Celldega
  • Cell type colocalization over the course of transplantation
  • Questions: How good is the pig genome?
  • Suggestions: Try out different technologies for adjacent slides?

MtDNA-based linage tracing in spatial data

  • Rong, Zhang, UPenn
  • mtDNA far more abundant in cells
  • Can linage information inferred from mtDNA implicate spatial changes?
  • Send cDNA to (1) spatial slide-seq (2) PCR amplification for genotyping
  • Input: MT Loci with cell type deconvolution (spots)
  • Question: Which variant is significantly over-representing cell types?
  • Assign MT variants to cell types
  • Barrett's Esophagus
  • Bracht and Rong et al. (2025)
  • Spot-level localization (testing the purity of nearest neighbors)

Can we simply use MT expression to construct cell lineage? Why not using expression-based CNV method?

Spatial neighbour analysis

  • Jun Inamo, U of Colorado
  • Juvenile Idiopathic Arthritis synovium
  • 10x Xenium, n=9
  • co-varying neighborhood analysis (Reshef YA, Nat Biotech 2022)
  • Can we relate this with locally-linear embedding?
  • Show dynamics of many fibroblast marker genes as a function of distance from endothelial cells
  • spatialCoocur (by permutation tests)
  • Compute colocalization scores between two cell types (e.g., T and myeloid, CXCL9-CXCR3)
  • Tertiary Lymphoid Structure (TLS): T-cell and B-cell together

Clonal hematopoiesis, non-coding putative driver mutations

  • Josh Weinstock, Emory
  • How does DNA changes over time?
  • Clonal hematopoiesis of indeterminate potential
  • Clonal extensions lack a mutation in canonical genes.
  • 490k UKBB whole genome
  • frequency of alternate allele as a function of age at blood draw
  • clonal hematopoiesis genes recapitulated!
  • AML risk was also captured in phewas on the variants
  • CH point mutations... how much variance was explained?
  • somatic.emory.edu
  • preprint
  • Questions from audience:
    • How do we know this is somatic or rare germline?
    • Most of them are germline point mutations.
    • Do cell types matter?
    • Copy number variants?
  • We may be able to ascertain "age" as a factor for somatic vs. germline... Does it matter?

After the ASHG meeting