Determining the Molecular Intermediates Between Genotype and Phenotype

on YouTube ↗25m

" summary

Uses pooled RIP-seq on nine iPSC lines plus genotype estimates and a Bayesian beta‑binomial model (implemented in Pyro) to call allele‑specific RBP binding; HNRNPAK shows ~7,000 ASB SNPs (~2% of testable sites) and ASB calls significantly enrich for fine‑mapped Leafcutter sQTLs. Develops a convex optimization/ADMM approach to infer directed causal networks from MR estimates that scales to hundreds of nodes; applied to ~400 UK Biobank traits it finds ~5% significant total causal effects, a direct network of ~3,000 edges, median path length two, and hubs like BMI.

AI-written summary · about

" chapters (7)

1. 0:00Start
2. 6:17Don't need many individuals to captur substantial common genetic variation
3. 8:24Deconvolution can estimate representation of each individual
4. 10:21A Bayesian probabilistic model distinguishes allele specific binding from expression
5. 16:28A linear model to capture network and genetic effects
6. 19:22Bidirectional mediated Mendelian randomization (bimmer)
7. 21:09biobank