Image-based epigenetic profiling with deep learning and high-speed super-resolution microscopy

BACKGROUND: Comprehensive profiling of epigenetic states is essential for understanding gene regulation and disease mechanisms. Sequencing-based methods such as ChIP-seq, Hi-C, and RNA-seq provide genome-wide views of histone modifications and 3D genome organization, but lack spatial resolution within single nuclei. RESULTS: Here we present an image-based epigenetic profiling framework that combines high-speed super-resolution microscopy with deep learning. Using models of (i) histone deacetylase inhibition in HEK293T cells and (ii) Rett syndrome iPS cells carrying MECP2 mutations, our approach accurately discriminated their epigenetic states (99.6% and 96.1% accuracy, respectively) and identified the nuclear periphery as a hotspot of H3K27ac and CTCF redistribution. Sequencing-based analyses showed compartment switching and lamina-associated domain alterations consistent with the image-based features. These results demonstrate that high-speed super-resolution imaging, when combined with deep learning, provides a powerful tool for epigenetic profiling. CONCLUSIONS: Our framework offers a generalizable strategy for image-based epigenetic profiling to uncover chromatin alterations in development, disease, and therapeutic response.