ChAI-Net learns a regulatory code for chromatin co-accessibility

ChAI-Net learns locus-specific chromatin co-accessibility maps from mouse embryonic stem cell DNA sequence, capturing unique enhancer, promoter, and CTCF-site motif combinations linked to co-accessibility.

• Why it matters: Bulk accessibility averages can miss coordinated protein-DNA interactions and nucleosome-position variation, leaving the regulatory “grammar” behind co-accessibility unclear at near-nucleotide resolution.
• What they did: The study quantifies pairwise simultaneous accessibility on single molecules from long-read footprinting, then trains ChAI-Net to predict co-accessibility maps from DNA sequence and tests learned rules via model interpretation and in silico experiments.
• The result: Interpretable motif orientation, spacing, positioning, and co-occurrence are highlighted as drivers of co-accessibility, and many motifs are missed by average-accessibility modeling, enabling extraction of regulatory rules for TF cooperativity.

AI summary of the abstract · may contain errors · not medical advice · read the paper.