HiCMamba: Enhancing Hi-C Resolution and Identifying 3D Genome Structures with State Space Modeling

2025-03-13

AI Summary

Hi-C data often suffer from low-resolution contact maps due to insufficient sequencing depth, hindering accurate 3D genome architecture inference. To address this, we propose HiCMamba—the first deep learning method for Hi-C super-resolution reconstruction that integrates state space models (SSMs) with a U-Net–style autoencoder and a novel multi-scale holistic scanning module. This design jointly captures long-range genomic dependencies and fine-grained local patterns. Crucially, the holistic scanning mechanism fuses multi-scale receptive fields while substantially reducing GPU memory consumption and inference latency. HiCMamba achieves state-of-the-art performance across multiple benchmarks. Reconstructed topologically associating domains (TADs) and chromatin loops exhibit strong concordance with epigenetic markers—including CTCF binding sites and histone modifications—demonstrating its biological fidelity and utility for downstream functional genomics analysis.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Computer Vision: Multi-modal Vision
• Search and Optimization: Metareasoning and Metaheuristics

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Hi-C technology measures genome-wide interaction frequencies, providing a powerful tool for studying the 3D genomic structure within the nucleus. However, high sequencing costs and technical challenges often result in Hi-C data with limited coverage, leading to imprecise estimates of chromatin interaction frequencies. To address this issue, we present a novel deep learning-based method HiCMamba to enhance the resolution of Hi-C contact maps using a state space model. We adopt the UNet-based auto-encoder architecture to stack the proposed holistic scan block, enabling the perception of both global and local receptive fields at multiple scales. Experimental results demonstrate that HiCMamba outperforms state-of-the-art methods while significantly reducing computational resources. Furthermore, the 3D genome structures, including topologically associating domains (TADs) and loops, identified in the contact maps recovered by HiCMamba are validated through associated epigenomic features. Our work demonstrates the potential of a state space model as foundational frameworks in the field of Hi-C resolution enhancement.

Problem

• Enhances Hi-C resolution using deep learning and state space modeling.
• Reduces computational resources while improving 3D genome structure identification.
• Validates recovered 3D genome structures with epigenomic features.

Innovation

• Deep learning enhances Hi-C resolution.
• State space model reduces computational resources.
• UNet-based auto-encoder perceives multi-scale fields.