Marcel Schulz - @[email protected]

Species‑aware DNA LMs just became practical. TransCodon learns 5′ UTR + CDS + RNA structure to rewrite genes with 49% native‑codon recall🆗 ...keeping rare pauses that guide folding and without CAI inflation🧬

nature.com/articles/s4158…

rdcu.be/d7rux Mapping the topography of spatial gene expression with interpretable deep learning

🔬 Toward histopathology 2.0: spatial transcriptomes inferred from routine diagnostic H&E images + a chat interface for cell-resolution histopathology through English language. (1/6)

The mainstream view of AI for science says AI will rapidly accelerate science, and that we're on track to cure cancer, double the human lifespan, colonize space, and achieve a century of progress in the next decade. In a new AI Snake Oil essay, @random_walker and I argue that…

Excited to share our @Nature paper on gut dysbiosis-related metabolite inducing atherosclerosis, led by @SanchoLab and a fantastic team!

Our work with the STATE model tackled context generalization for *seen* perturbations. @arcinstitute's Virtual Cell Challenge (VCC) focuses on a more challenging scenario: context generalization for *unseen* perturbations. Colab notebook to train your own STATE models for VCC:…

1/10 ML can solve PDEs – but precision🔬is still a challenge. Towards high-precision methods for scientific problems, we introduce BWLer 🎳, a new architecture for physics-informed learning achieving (near-)machine-precision (up to 10⁻¹² RMSE) on benchmark PDEs. 🧵How it works:

🔥It is a pleasure to announce CrossTalker2.0🔥. New features includes CCI statistical tests, new visualizations, intracellular signalling, pathway analysis, a python version & tutorials based on heart and bone marrow fibrosis data costalab.github.io/CrossTalkeR/ by Vanessa & @shin_nagai

More than 2700 3′UTRs are highly conserved. These 3′UTRs are essential components in mRNA templates, as their deletion decreases protein activity without changing protein abundance. Highly conserved 3′UTRs help the folding of proteins with long IDRs. biorxiv.org/content/10.110…

"Here we present the first direct evidence that H3K4me3 is not a cause but a consequence of transcriptional activation, functioning as a downstream epigenetic response" biorxiv.org/content/10.110…

Coordinated regulation by lncRNAs results in tight lncRNA-target couplings dlvr.it/TLnKgm

The human transcription factor occupancy landscape viewed using high-resolution in situ base-conversion strand-specific ... biorxiv.org/content/10.110… #biorxiv_genomic

🚀 What do genomic Transformers actually learn about biology? •What knowledge do they hold at random init, after pre‑training, and following fine‑tuning? •We dove deep into every attention head to find out. 📄 Preprint live now “Interpreting Attention Mechanisms in Genomic…

It really does nothing of the sort. It’s so well researched and understood that in 2010 when it was published in that obscure science journal Nature, I commissioned a Lego version and made a film about it. youtu.be/RLPVCJjTNgk?si…

It is finally out! My first @ScienceMagazine paper is online and it does not seem real😃. I am Incredibly happy! Really thanks to all my coauthors and special mention to my incredible supervisor @iDikic2 for all the support and trust ! 🫶. Science science.org/doi/10.1126/sc…

Excited to share our study on spatio-temporal reconstruction of cardiac scar formation led by @ChanAndyson. We combine scRNA-seq with spatial transcriptomics to resolve the fibrotic niche and the niche of dedifferentiating cardiomyocytes. biorxiv.org/content/10.110…

Wow nature.com/articles/s4158…

Check out our new work led by @ashton_omdahl discussing matrix factorization across GWAS, accounting for sample overlap. Factors with varying polygenicity, enrichment for cell type / developmental stage... Ashton is speaking at #biodata24 today! biorxiv.org/content/10.110…

check out our new preprint led by @Charles_Zhou12 and supervised by @MengjieChen6 and me doi.org/10.1101/2024.1… where we present scPrediXcan which integrates deep learning and single cell expression data into a powerful cell type specific TWAS framework