SMFNet: a diffusion model fits single-molecule chromatin state distributions at 2Kb context
anshulkundaje · x · 2026-11-06
Anshul Kundaje's team (EMBL Heidelberg + Stanford) posted a bioRxiv paper, SMFNet reveals principles of combinatorial transcription factor assembly, introducing a sequence-conditioned generative diffusion model that fits chromatin state distributions from single-molecule footprinting data (SMF, Fiber-seq, etc.), currently at 2 Kb context.
The authors note that prior models — BPNet, ChromBPNet, ProCapNet — achieved single base-pair resolution and demonstrated the value of model interpretation for regulatory syntax, but were trained on profiles averaged over many cells and molecules, hiding the heterogeneity and dynamics of regulatory states within the same cell type. SMFNet moves to single-molecule resolution and models the full distribution of chromatin states.
The team also presents PRISM, a single-molecule in silico mutagenesis approach for probing sequence features (motifs and syntax) that affect single-molecule chromatin states, with in silico syntax and variant perturbation experiments validating well. The authors expect these models to unify ML and biophysics for gene regulation and enable exploration of underlying biophysics through a sequence-conditioned generative lens.
More from Research
- Caltech's Van Valen Lab ditches X, publishes all AI models on Hugging Face — anshulkundaje · 2026-11-06
- Soumith Chintala: Inkling 2 base model beats GLM-5.3, DeepSeek V4 on code — soumithchintala · 2026-11-06
- Cartesia's H-Net shows in-context tokenization: models can learn to chunk data without a tokenizer — TheTuringPost · 2026-11-06
- Independent audit: RoundSquare's zero-axiom Lean 4 proof compiles cleanly but doesn't resolve Clay Yang–Mills — kevrussell · 2026-11-06
- Learning-theoretic protocol lets two companies privately test data compatibility before pooling — roydanroy · 2026-11-06
- AI-run lab completes 11,000+ autonomous experiments; Claude parses 15B protein sequences — karlwaldman · 2026-11-06