COCOA: A Framework for Fine-scale Mapping Cell-type-specific Chromatin Compartmentalization Using Epigenomic Information

Abstract Chromatin compartmentalization and epigenomic modification are crucial factors in cell differentiation and diseases development. However, mapping precise chromatin compartmental patterns across multiple cell types requires Hi-C or Micro-C data at high sequencing depth. Exploring the systematic relationship between epigenomic modifications and compartmental patterns remains a challenge. To address these issues, we present COCOA, a deep neural network framework that uses convolution and attention mechanisms to infer reliable fine-scale chromatin compartment patterns from six representative histone modification signals. COCOA achieves this by extracting 1-D track features through bi-directional feature reconstruction after resolution-specific binning epigenomic signals. These track features are then cross-fused with contact features using an attention mechanism. Subsequently, the contact features are transformed into chromatin compartment patterns through residual feature reduction. COCOA demonstrates accurate inference of chromatin compartmentalization at a fine-scale resolution and exhibits stable performance on test sets. In addition, we explored the impact of histone modifications on the chromatin compartmentalization through in silico epigenomic perturbation experiments. When using 1kb resolution high-depth experimental data, obscure compartments are observed, whereas COCOA can generate clear and detailed compartmental patterns. Finally, we demonstrated that COCOA enables cell-type-specific prediction of unrevealed chromatin compartment patterns in various biological processes. Thus, COCOA is an effective tool for gaining chromatin compartmentalization insights from epigenomics in a wide range of biological scenarios.