Related Experiment Video
Updated: Jul 4, 2026
06:07
Continuous Fluorescence-Based Endonuclease-Coupled DNA Methylation Assay to Screen for DNA Methyltransferase Inhibitors
Published on: August 5, 2022
CMA-Nano: A DNA Methylation Detection Method for Nanopore Sequencing Data Based on a Cross-Modal Attention Mechanism
Hua Shi1, Chenglin Yin2, Jianbo Qiao2
1School of Optoelectronic and Communication Engineering, Xiamen University of Technology, Xiamen 361024, China.
ACS Omega
|July 3, 2026
Summary
CMA-Nano, a deep learning tool, accurately detects DNA methylation from nanopore sequencing data by integrating multimodal features. This method enhances gene regulation and disease research by overcoming signal noise challenges.
Area of Science:
- Genomics
- Bioinformatics
- Molecular Biology
Background:
- DNA methylation (5-methylcytosine) is crucial for gene regulation, development, and disease.
- Nanopore sequencing offers direct DNA methylation detection but faces challenges due to signal noise and complex data.
Purpose of the Study:
- To develop CMA-Nano, a deep learning framework for accurate DNA methylation detection using nanopore sequencing.
- To address analytical challenges posed by signal noise and multimodal data in nanopore sequencing.
Main Methods:
- Utilized a cross-modal attention mechanism to integrate DNA sequences, nanopore signal statistics, and raw current signals.
- Implemented a masked base prediction pretraining strategy for enhanced feature representation and generalization.
- Evaluated performance on diverse datasets including *Arabidopsis thaliana*, *Oryza sativa*, and human samples.
Main Results:
- CMA-Nano demonstrated superior performance compared to existing methods across key metrics.
- Ablation studies confirmed the effectiveness of cross-modal attention and multimodal fusion.
- Transfer learning and interpretability analyses highlighted the model's predictive accuracy and biological relevance.
Conclusions:
- CMA-Nano provides a robust and generalizable solution for DNA methylation detection from nanopore sequencing data.
- The framework effectively fuses multimodal features, capturing complex interdependencies for improved accuracy.
