Related Experiment Video
Updated: Jan 6, 2026
09:47
Author Spotlight: Advancing Alzheimer's Research – Exploring Early Detection and Multi-Omics Approaches
Published on: December 15, 2023
1.6K
IMATAC imputes single-cell ATAC-seq data by deep hierarchical network with denoising autoencoder.
Yao Li1, Hongqiang Lyu1, Kexin Li1
1Faculty of Electronic and Information Engineering, School of Automation Science and Engineering, Xi'an Jiaotong University, No. 28 Xianning West Road, Beilin District, Xi'an, Shaanxi 710049, China.
Briefings in Bioinformatics
|September 29, 2025
Summary
IMATAC, a novel deep learning method, effectively imputes missing data in single-cell ATAC-seq (scATAC-seq) by addressing dropout events. This improves downstream analyses like cell clustering and regulatory element discovery.
Area of Science:
- Genomics
- Computational Biology
- Bioinformatics
Background:
- Single-cell ATAC-seq (scATAC-seq) enables chromatin accessibility profiling at the individual cell level.
- Dropout events, where true signals are missed, are a significant challenge in scATAC-seq data, hindering downstream analysis.
- Imputing sparse, high-dimensional scATAC-seq data remains difficult due to its unique properties.
Purpose of the Study:
- To develop an effective imputation method for scATAC-seq data to overcome the limitations imposed by dropout events.
- To introduce IMATAC, a deep hierarchical network with a denoising autoencoder, for accurate scATAC-seq data imputation.
- To enhance downstream analyses by improving the quality of scATAC-seq data through imputation.
Main Methods:
- Proposed IMATAC, a deep hierarchical network incorporating a denoising autoencoder for scATAC-seq data imputation.
- Embedded scATAC-seq data into a latent space using a two-level hierarchical architecture (local details and global information).
- Utilized a denoising autoencoder and a parallel multi-classifier to reconstruct original data and recover missing values within cell populations.
Main Results:
- IMATAC demonstrated superior performance compared to existing methods in imputation accuracy on both simulated and experimental data.
- The method achieved lower imputation errors, effectively distinguishing dropout zeros from biological zeros.
- Improved downstream analyses, including heterogeneous clustering, differential analysis, and regulatory element discovery.
Conclusions:
- IMATAC provides an effective solution for imputing sparse scATAC-seq data, significantly mitigating the impact of dropout events.
- The deep hierarchical network architecture and denoising autoencoder contribute to the method's ability to capture complex data structures.
- IMATAC enhances the reliability and utility of scATAC-seq data for biological discovery.
