SpaDC enables sequence-based integrative analysis and regulatory inference of spatial chromatin accessibility data

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Spatial ATAC-seq offers simultaneous profiling of cellular locations and chromatin accessibility in intact tissues.
• Existing methods face challenges like high dimensionality, noise, sparsity, and often overlook crucial DNA sequence information containing regulatory motifs.

Purpose of the Study:

• To introduce SpaDC, a novel graph-regularized convolutional neural network designed to overcome limitations in spatial ATAC-seq analysis.
• To integrate spatial location, chromatin accessibility, and DNA sequence data for enhanced analytical power.
• To improve the accuracy and robustness of spatial ATAC-seq data interpretation.

Main Methods:

• Developed SpaDC, a graph-regularized convolutional neural network.
• Utilized a triplet loss function for integrating multiple spatial ATAC-seq datasets and mitigating batch effects.
• Performed benchmark analyses on real datasets to evaluate performance.

Main Results:

• SpaDC demonstrated state-of-the-art performance in spatial domain identification, data denoising, and gene regulatory network (GRN) inference.
• Applied to mouse embryonic brain data, SpaDC accurately identified known brain structures and recovered chromatin accessibility signals.
• Analysis of P22 mouse brain spatial multi-omics data revealed spatial domain-specific cis-regulatory elements and GRNs.

Conclusions:

• SpaDC provides a powerful, sequence-based solution for spatial ATAC-seq analysis.
• The method enables more accurate and robust investigation of tissue architecture and chromatin organization.
• SpaDC advances the field by incorporating DNA sequence information into spatial multi-omics analyses.