DNA-BFN: Bayesian Flow Networks for Generative Modeling of Regulatory DNA Sequences.

DNA Bayesian Flow Networks (DNA-BFN) outperform state-of-the-art models in generating regulatory DNA sequences, consistently across eight diverse datasets, with improved fidelity and efficiency.

• Why it matters: Accurate design of regulatory DNA sequences is hindered by their complex dependencies and the limitations of existing generative models that can distort sequence structure, impeding biological relevance and practical applications.
• What they did: The authors developed DNA-BFN, a framework that models nucleotide distributions through iterative Bayesian updates, enabling diffusion-like sequence generation without disrupting discrete sequence structure, trained on eight datasets.
• The result: DNA-BFN faithfully preserves sequence features, supports few-shot design, and offers faster sampling, demonstrating a generalizable approach for nucleotide sequence generation and design.