I am excited about all three strands. Work from @oliverstegle.bsky.social on things like DeepRVat and my own lab on Delphi-2M are methodological innovations for biological methods; Anna Kreshuk has been leading the field in image analysis with Ilastik etc and

#ESHG attendees: Hope to see you at my presentation about our Nature Genetics paper on deep learning for rare variant genetics. Sunday 8:30-9:15am in session E02.

It's an educational session for non-experts in AI, so I'll start from the basics and explain the principles behind the model in detail.

Why be satisfied with coding variant gene impairments? Check out Eva’s poster to learn how we scale DeepRVAT to the WGS UKBiobank. It’s been great fun being a part of this project!
#ASHG25

@thbec.bsky.social is going to share preliminary results on Meta-DeepRVAT, a new approach for deep learning based meta-analysis improving the power of rare variant association studies using population scale external control cohorts. #MLCSB
📅 July 21 |📍 Poster A-312

@thbec.bsky.social presents his latest work on DeepRVAT. He showcases how deep learning-based meta-analysis can improve power for genetic association studies in rare diseases using biobank-scale data — all while preserving privacy.
📍 Poster PSII-37 | 🕓 Wed 16:00

@brianfclarke.bsky.social and @thbec.bsky.social will present their latest work at the @helmholtzai.bsky.social conference in Karlsruhe this week!

Two exciting posters on rare variant genetics — both on Wednesday at 16:00. Details below 👇

#HelmholtzAI #Genomics #RareVariants #DeepRVAT #HAICON

We used multiple approaches to check whether these associations are real. We ran FlexRV in 200k UKB individuals and checked whether these replicated in the reported DeepRVAT results on the full cohort - and found that they replicated at a higher rate than Regenie or STAAR.

For example, DeepRVAT is a very cool deep learning method that combines many variant annotations together into a “gene impairment score” for each individual. But for the 28 quantitative traits we tested in UKB, FlexRV found 37% more associations (and 58% more for binary)!

#eshg2025 SL FuncRVP is their new method, compared to DeepRVAT it is better in 38/41 traits in UKB - best improvment for more genetic traits or ones with with more GWAS hits. - method works for rare and common vars - outperforms other modesl - gene expression data adds the most info

#eshg2025 now computational session: Shuubhankar Londhe using gene embeddings to improve rare variant PRSs- dev'd DeepRVAT a rare variant associated test; gene embeddings capture func relationships between genes as numerical vectors - PoPS and other method have done this w GWAS. What about rare?

I might be biased, but would strongly recommend! Go and checkout how you can use DeepRVAT in your research + some teasers on what v2 will hold at #ESHG 🧬🔎

DeepRVATv2 is in the making! We learn from >60M WGS variants, >90 features, and 500k UKB participants to predict gene impairment beyond coding regions and showcase how DeepRVAT can advance rare disease diagnostics and gene constraint measurement.

Learn more at my Poster 5058W, Wed 2:30pm #ASHG25

Ever wondered how to choose weights for rare variant burden/SKAT tests? @jeremy37.bsky.social and I did too! Check out our (@illumina) straightforward new approach that improves discovery by ~50% or more than STAAR, Regenie, and DeepRVAT!

FuncRVP significantly improves phenotype prediction compared to linear models using DeepRVAT gene impairment scores (pg. 9). Figure 3 demonstrates this, showing superior performance for 25/41 traits, especially those with high burden heritability, numerous gene associations, and many significant...