minja

When I created Twitter account, I promised myself not to post anything unrelated to science. However, the current situation goes beyond all conceivable limits. Unfortunately, I can't change anything, but at least I can say it here: THE WAR MUST BE STOPPED.

Currently, we have only preliminary benchmarks on @instadeepai NT bench & our own gene finding bench (iclr.cc/virtual/2026/1…). Looking forward to more thorough benchmarking from the community! @volokuleshov @anshulkundaje @exnx @pkoo562 @BiologyAIDaily

In our paper (biorxiv.org/content/10.648…) we study how modern Transformer improvements transfer to genomics. We show that the ModernBERT-based architecture behind ModernGENA achieves the high throughput while maintaining competitive performance.

We will present our new DNA Language model, ModernGENA (huggingface.co/collections/AI…), at #ICLR workshops on Monday, April 27! Our posters will be presented at #ICLR workshops on Monday, April 27: FM4Science 4:00 PM MLGenX 3:40 PM Gen2 4:10 PM We welcome discussions on DNA LMs!

I think it aligns well with what @dekker_lab @golobor , and others showed before for other cell types - but in LBC, hypertranscription results in an extreme, amplified version of these general principles

8/ A key insight: 👉 Gene orientation matters Convergent genes (→ ←): → trap cohesin between them → form strong insulation boundaries This explains domain formation without CTCF

1/ Our paper just came out in Nucleic Acids Research! 🎉 We study one of the most fascinating chromatin systems — lampbrush chromosomes (LBCs) in chicken oocytes. 🔗 academic.oup.com/nar/article/54…

10/ Huge thanks to collaborators from Saint-Peterburg University (Alla Krasikova’s group), experimental (Maria Gridina and co) & bioinf/biophys (Timofey Lagunov and co) teams! This was a great pleasure working with them!

9/ We tested this with polymer simulations Model includes: • loop extrusion • RNA polymerase as moving barrier • cohesin repositioning • RNP-driven repulsion Result: ✅ recapitulates Hi-C maps ✅ reproduces loop/chromomere structure

Our tool Charm is out in @NAR_Open Genomics & Bioinformatics! Simulate Hi-C data for CNV, inversions, translocations & complex SV. academic.oup.com/nargab/article… 💻 github.com/genomech/charm #HiC #Genomics #SV #Bioinformatics #CharmTool

It's common knowledge that H3K9me3 correlates with DNA methylation. Can anyone point me to public human datasets of H3K9me3 ChIP-seq & WGBS that I can load to IGV to validate this correlation visually? Tried with K562 data from Encode, can't see any correlation.

For wet-lab scientists, our user-friendly web service lets you analyze DNA sequences using specialized instruments based on GENA-LMs with just a click at dnalm.airi.net No bioinformatics skills are required!

Interested in trying GENA? If you're an ML expert or bioinformatician, start with our GitHub (github.com/AIRI-Institute…) or HuggingFace (huggingface.co/collections/AI…) repos.

GENA-LMs is not a single model - it's a collection of models, including taxon-specific pretrainings, optimized for human, fly, yeast, or Arabidopsis. These specialized models enhance performance in species-specific genomic tasks. 7/n

More on long context: GENA's recurrent memory extends context to >100kb. This technique breaks the existing trade-off between model size and input lengths, allowing the application of large models on long inputs. More details in our ICM workshop paper: openreview.net/forum?id=K671l… 6/n

Our DNA Language Model 🧬GENA🧬 is now published in @NAR_Oxford! Read the paper (academic.oup.com/nar/article/53…) & access examples on GitHub (github.com/AIRI-Institute…) or HuggingFace (huggingface.co/collections/AI…). Thread below 🧵1/n