Tensor decomposition discriminates tissues using scATAC-seq.

Area of Science:

• Genomics
• Computational Biology
• Single-cell Analysis

Background:

• Assay for Transposase-Accessible Chromatin using sequencing (ATAC-seq) maps chromosome structure.
• Single-cell ATAC-seq (scATAC-seq) offers single-cell resolution but suffers from data sparsity.
• Existing methods struggle with the large, sparse matrices generated by scATAC-seq.

Purpose of the Study:

• To introduce tensor decomposition (TD) as a method for imputing missing values in scATAC-seq data.
• To demonstrate the utility of TD for analyzing massive scATAC-seq datasets.
• To overcome the limitations of data sparsity in single-cell chromatin accessibility studies.

Main Methods:

• Application of tensor decomposition (TD) to large-scale scATAC-seq datasets.
• Processing genomic intervals of approximately 200 bp, potentially reaching over 13 million intervals.
• Comparison with existing methods for handling large sparse matrices.

Main Results:

• TD successfully filled missing values in massive scATAC-seq datasets.
• Generated UMAP embeddings that reflect tissue specificity.
• Identified genes associated with biological enrichment terms and transcription factor targeting.
• Demonstrated the capability to handle datasets with over 13 million genomic intervals.

Conclusions:

• Tensor decomposition is a powerful tool for processing large, sparse scATAC-seq data.
• TD facilitates the extraction of meaningful biological insights from sparse single-cell chromatin accessibility data.
• The proposed TD method overcomes limitations of current approaches for scATAC-seq data analysis.