SCAN-ATAC

SCAN-ATAC simulates single-cell ATAC-seq (scATAC-seq) data from bulk-tissue ATAC-seq experiments to produce benchmark datasets with known cell-type labels for evaluating clustering, deconvolution, and other scATAC-seq analyses.

Key Features:

• Simulation methodology: Down-samples bulk ATAC-seq data from representative cell lines or tissues to create simulated scATAC-seq experiments.
• Tunable signal-to-noise ratio: Incorporates a consistent but adjustable signal-to-noise ratio across cell types to integrate bulk experiments with varying background noise.
• Diploid genome consideration: Accounts for diploid genomes by independently sampling twice without replacement to reflect two genomic copies.
• Sampling algorithm: Uses an efficient weighted reservoir sampling algorithm for selection of reads or fragments.
• Performance and scalability: Implemented in C++ with OpenMP parallelization to enable rapid simulation of millions of cells in less than an hour on a standard laptop.

Scientific Applications:

• Benchmarking scATAC-seq analysis: Provides simulated datasets with known cell-type labels for validating clustering and deconvolution methods.
• Method evaluation: Enables rigorous testing of clustering algorithms and deconvolution approaches under controlled signal-to-noise conditions.
• Regulatory landscape characterization: Supports studies aiming to assess cell-type-specific regulatory element detection and interpretation in single-cell epigenomics.

Methodology:

Computational steps explicitly include down-sampling bulk ATAC-seq data, applying a tunable signal-to-noise model, independently sampling twice without replacement for diploid genomes, using a weighted reservoir sampling algorithm, and implementation in C++ with OpenMP parallelization.