A generalizable Hi-C foundation model for chromatin architecture, single-cell and multiomics analysis across species

2026-06-15 · Source: Machine learning : nature.com subject feeds · Field: Science & Research — Life Sciences & Biology, Artificial Intelligence & Machine Learning, Research Methodology & Innovation · Depth: Expert, extended

Summary

HiCFoundation is a novel foundation model designed for the integrative analysis of three-dimensional (3D) chromatin structure, measured by high-throughput chromosome conformation capture (Hi-C), and associated epigenomic regulation, such as assay for transposase-accessible chromatin using sequencing (ATAC-seq) and chromatin immunoprecipitation followed by sequencing (ChIP-seq). Trained on extensive Hi-C data, this model achieves high performance and generalizability across species for various 3D genome analyses, including reproducibility analysis, resolution enhancement, and loop detection. HiCFoundation can also predict diverse epigenomic activities directly from Hi-C data, elucidating the link between 3D structure and regulatory function. Furthermore, the model adapts easily to single-cell Hi-C data, offering a versatile and interpretable framework for studying the 3D genome and its functional roles across different cell types and species. The code is available on GitHub and Zenodo with an Apache 2.0 license.

Key takeaway

For genomics researchers analyzing complex chromatin structures, you should consider integrating HiCFoundation into your workflow to overcome challenges in combining 3D structure and epigenomic data. This model provides a unified, interpretable framework for multi-species analysis, improving reproducibility, resolution, and loop detection. Its ability to predict epigenomic activities from Hi-C data can accelerate your understanding of regulatory functions. Explore the GitHub repository for implementation.

Key insights

HiCFoundation integrates 3D chromatin structure and epigenomic data across species using a generalizable foundation model.

Principles

• Foundation models enhance generalizability in genomics research.
• 3D genome structure directly links to regulatory function.
• Single-cell Hi-C analysis benefits from specialized model adaptation.

Method

HiCFoundation is pre-trained on massive Hi-C data, then fine-tuned for specific tasks like reproducibility analysis, resolution enhancement, loop detection, and epigenomic activity prediction.

In practice

• Use HiCFoundation for cross-species 3D genome analysis.
• Predict epigenomic activities directly from Hi-C data.
• Apply to single-cell Hi-C for enhanced resolution.

Topics

• Hi-C Data Analysis
• Foundation Models
• Chromatin Architecture
• Epigenomics
• Single-cell Genomics
• Multiomics Integration

Code references

• Noble-Lab/HiCFoundation_paper
• Noble-Lab/HiCFoundation
• Noble-Lab/HiCFoundation

Best for: AI Scientist, Research Scientist, Machine Learning Engineer

Related on AIssential

• Prediction and functional interpretation of inter-chromosomal genome architecture from DNA sequence with TwinC — TwinC predicts inter-chromosomal DNA contacts, revealing their mechanistic drivers and role in gene regulation.
• Machine learning framework reveals a concordant cell-state landscape across single-cell datasets — CONCORD is a contrastive learning framework for integrating single-cell omics data to reveal cell-state landscapes.
• Rethinking Genomic Modeling Through Optical Character Recognition — OpticalDNA reframes genomic modeling as OCR-style document understanding, achieving efficiency and high-fidelity compression via visual layouts.
• Exascale Multi-Task Graph Foundation Models for Imbalanced, Multi-Fidelity Atomistic Data — Exascale graph foundation models accelerate materials discovery by screening billions of structures in seconds.
• Biological Reasoning-Informed Regression for Interpretable Regulatory DNA Activity Prediction — Integrating structured biological knowledge with LLMs via two-stage training improves interpretable DNA regulatory activity prediction.
• Carbon, open source DNA model, 250x faster than Evo2-7B and runs on llama.cpp — Applying LLM architectures to DNA sequences enables highly efficient genomic analysis and structure prediction.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine learning : nature.com subject feeds.