scaDA: A Novel Statistical Method for Differential Analysis of Single-Cell Chromatin Accessibility Sequencing Data

Area of Science:

• Genomics
• Computational Biology
• Statistical Genetics

Background:

• Single-cell ATAC-seq (scATAC-seq) is crucial for studying chromatin accessibility at a single-cell resolution.
• Differential chromatin accessibility (DA) analysis is a key application, but faces challenges due to scATAC-seq data's high zero counts and variability.
• Current DA methods often focus on mean differences, neglecting important distributional variations observed in real data.

Approach:

• Introduced scaDA, a novel composite statistical test utilizing a zero-inflated negative binomial (ZINB) model.
• scaDA performs differential distribution analysis by simultaneously testing chromatin accessibility abundance, prevalence, and dispersion.
• The method incorporates dispersion shrinkage and iterative refinement for robust parameter estimation.

Key Points:

• scaDA demonstrated superior performance over existing ZINB-based likelihood ratio tests and other published methods in simulations.
• Achieved highest power and best false discovery rate (FDR) control in comprehensive simulation studies.
• Outperformed other methods in three real single-cell multi-omics datasets.

Conclusions:

• scaDA successfully identified differentially accessible regions in microglia from single-cell multi-omics data for an Alzheimer's disease (AD) study.
• Identified regions were significantly enriched in Gene Ontology (GO) terms related to neurogenesis, a known AD phenotype.
• Results highlight scaDA's utility in uncovering biologically relevant genomic features associated with complex diseases like AD.