epiConv

epiConv integrates single-cell ATAC-seq (scATAC-seq) datasets to correct batch effects and enable joint chromatin accessibility analyses across diverse biological conditions.

Key Features:

• Batch Effect Correction: Mitigates batch effects by distinguishing technical noise from genuine biological variation in scATAC-seq datasets.
• Resistance to Over-Fitting: Demonstrates reduced propensity for over-fitting when integrating complex datasets such as peripheral blood mononuclear cells (PBMCs).
• Alignment and Resolution Enhancement: Aligns low-depth scATAC-seq data from co-assay experiments (joint transcriptome and chromatin profiling) to high-quality ATAC-seq references to enhance chromatin profile resolution.
• Integration Across Biological Conditions: Integrates cells across conditions (e.g., T cells in normal versus germ-free mice, normal versus malignant hematopoiesis) to reveal cell populations hidden in separate analyses.
• Improved Clustering and Peak Calling: Enables joint analyses that improve clustering performance and differentially accessible peak calling when biological signals are weak in individual datasets.

Scientific Applications:

• Chromatin dynamics and gene regulation: Investigating chromatin accessibility dynamics and gene regulation at single-cell resolution across cell types and conditions.
• Immune and PBMC studies: Analyzing peripheral blood mononuclear cells (PBMCs) and T cell states, including comparisons between conventional and germ-free mice.
• Hematopoiesis and cancer research: Comparing normal and malignant hematopoiesis to reveal disease-associated chromatin states and cell populations.
• Co-assay and low-depth data enhancement: Enhancing resolution and revealing hidden cellular populations in co-assay (transcriptome + chromatin) and low-depth scATAC-seq datasets.

Methodology:

Computational steps include distinguishing technical noise from biological variation for batch-effect correction, aligning low-depth co-assay scATAC-seq to high-quality ATAC-seq references to enhance resolution, and performing joint analyses to improve clustering and differentially accessible peak calling while reducing over-fitting.