iiinova — where your digital self keeps evolving

Chang M. Yun

@chang-m-yun.bsky.social

PhD Candidate @stanford.edu | @anshulkundaje.bsky.social Lab DNA 🧬, proteins 🧪, and engineering 🤖. Also into oil painting & craft beers. (prev. @columbiauniversity.bsky.social, @cam.ac.uk)

Posts Replies Media

Chang M. Yun @chang-m-yun.bsky.social · 09/09/2026

Huge congrats to the AlphaGenome team on AG Atlas! Great to see that MotifCompendium could help manage a *whopping* ~900M seq patterns via GPU pairwise similarity+cluster+annotation. V rewarding we could support such a meaningful resource. Real use case of AI actually helping solve disease. ✅

110

Chang M. Yun @chang-m-yun.bsky.social · 17/08/2026

ENCODE GRAMMAR is officially on UCSC @genomebrowser.bsky.social as a Public Track Hub! 🎉 Easiest way to view any of ENCODE GRAMMAR's 3,865 models x Experimentally observed profiles, model-predicted profiles, model-importance scores, and motif instance tracks: genome.ucsc.edu/s/Lou/ENCODE...

genome.ucsc.edu

031

Chang M. Yun @chang-m-yun.bsky.social · 10/08/2026

So what can you do with ENCODE GRAMMAR? While we’ve shown the ability of one s2f model, ENCODE GRAMMAR brings a new dimension, layering multiple models across cell type & modality. To illustrate, we showcase an example locus: MYC (also our main figure in the ENCODE 4 preprint) 1/

152

Reposted by Chang M. Yun

Selin Jessa @selinjessa.com · 06/08/2026

Got to make a small contribution to this careful work to assemble and also label (no small feat!!!) thousands of motifs from across hundreds of cell states, with a new, scalable, gpu-accelerated toolkit from @chang-m-yun.bsky.social and Salil to analyse and compare these types of motif compendia

152

Reposted by Chang M. Yun

Selin Jessa @selinjessa.com · 06/08/2026

Massive effort from many in @anshulkundaje.bsky.social lab to train and curate this collection of 1000s of sequence-to-function deep learning models and the predictive features they learn - check out their blog posts on how to use the resource!

083

Chang M. Yun @chang-m-yun.bsky.social · 06/08/2026

Finally… Excited to share what I’ve been working on for the first half of my PhD! ENCODE GRAMMAR: One of the largest collections of regulatory DNA seq2func models to date (3,865 in total) trained across ENCODE, with full model interpretations and annotations tracks. Blog: tinyurl.com/3hmknhaa 1/

13813