Denoising spatial epigenomic data via deep matrix factorization.

Area of Science:

• Genomics
• Computational Biology
• Epigenetics

Background:

• Spatial epigenomics (SE) technologies map epigenomic landscapes in intact tissues, preserving spatial context for in situ gene regulatory studies.
• Current SE datasets often exhibit low signal detection, high noise, and sparse peak matrices, hindering downstream analysis.

Purpose of the Study:

• To introduce SPEED (spatial epigenomic data denoising), a novel deep matrix factorization framework.
• To impute and denoise SE data by integrating atlas-level single-cell epigenomic data and spatial context.

Main Methods:

• Developed a deep matrix factorization framework named SPEED.
• Leveraged atlas-level single-cell epigenomic data and spatial context for data imputation and denoising.
• Benchmarked SPEED against five state-of-the-art methods on simulated and real SE tissue datasets.

Main Results:

• SPEED demonstrated superior performance over existing methods across diverse tissues and SE technologies.
• Denoised SE data from SPEED facilitated enhanced downstream analyses.
• Improved differential chromatin accessibility analysis, epigenomic spatial domain identification, and gene activity inference were observed.

Conclusions:

• SPEED is a generalizable tool for improving the quality of spatial epigenomic data.
• The framework enhances biological insights derived from SE datasets.
• SPEED offers a robust solution for challenges in spatial epigenomics data analysis.