Decoding Cell-type Specific Gene Regulation with CellReg3D by Integrating DNA Large Language Models and 3D Genome Structure

CellReg3D integrates DNA language model sequence encoding with Hi-C 3D chromatin interactions to learn cell-type-specific regulatory mechanisms, achieving state-of-the-art prediction of epigenetic marks and gene expression across multiple cell types.

• Why it matters: Capturing long-range genomic interactions and combining sequence features with chromatin contact data remains a central challenge for modelling gene regulation in a cell-type-specific way.
• What they did: The multiscale, multi-modal framework uses a DNA language model for sequence-level information and incorporates Hi-C-derived 3D chromatin interactions, then applies interpretable analyses to link learned patterns to regulatory mechanisms and variant effects.
• The result: It predicts epigenetic marks and gene expression profiles across multiple cell types at state-of-the-art performance, improves interpretation of eQTLs, and prioritizes functional non-coding variants and potential gene regulatory interactions.

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