Paired Insertion Counting for Single-Nucleus ATAC-Seq Data • PICsnATAC

The goal of PICsnATAC is to accurately quantify snATAC-seq data. The two key components are:

1. Constructing cell-by-peak matrix with Paired Insertion Counting (PIC).

Motivation:

There are two common counting approaches for snATAC-seq data, fragment-based counting and insertion-based counting. They generate inconsistent results (see figure below).

Here, we present PIC as a uniform counting choice for various assays, including 10X Genomics snATAC-seq, sci-ATAC-seq, SNARE-seq, dsc-ATAC-seq, s3-ATAC-seq, and scNANO-ATAC.

2. Conducting Differential Accessible Region (DAR) identification analysis using the PIC matrix.

Motivation:

Two insertion events with correct primer configurations are required for one viable fragment. We derived the probability distribution of fragment count as a function of insertion rate and derived a model-based insertion rate estimation framework. We further applied a likelihood ratio test approach to detect DARs between populations.

Building upon our model-based insertion rate estimation framework, this analysis provides a more sensitive test of DAR between cell groups (cell types).

Dependencies

Please install the dependent libraries by running the following codes

install.packages('data.table') ## (please make sure it is newer than 1.8)
if (!require("BiocManager", quietly = TRUE))
    install.packages("BiocManager")
BiocManager::install(c('IRanges', 'GenomicRanges','Rsamtools', 'GenomeInfoDb'))

Installation

You can install the most up-to-date version of PICsnATAC from GitHub with:

# install.packages("devtools")
devtools::install_github("Zhen-Miao/PICsnATAC")

PIC-snATAC is a light-weight package. Installation typically takes <5 min.

Example

This is a basic example which shows you how to construct PIC-based cell by peak matrix:

library(PICsnATAC)

pic_matrix <- PIC_counting(
  cells = cells,
  fragment_tsv_gz_file_location = fragment_tsv_gz_file_location,
  peak_sets = peak_sets
)

DAR_by_LRT() treats counts strictly greater than its configurable artifact_threshold (default: 10) as unreliable and replaces them with zero before capping the remaining counts at the model-supported value of 5. The threshold is a heuristic for possible mapping errors or unusual fragment structures; use artifact_threshold = Inf to disable it and consider a sensitivity analysis when high counts are common in your assay.

Vignettes

The vignettes can be found at our pkgdown page:

https://zhen-miao.github.io/PICsnATAC/

Interactive Notebooks

Here, we provided two Google Colab notebooks for users to run the codes on cloud (processed example data and link provided). However, as snATAC-seq data are typically large and computer intensive, it is recommended to run it using computing cluster for large datasets

Notebook 1: Run PIC counting on PBMC data.

Notebook 2: Run PIC DAR on SNARE-seq2 brain data.

Citation

If you use PICsnATAC in your research, please cite our paper:

Miao Z, and Kim J. Uniform quantification of single-nucleus ATAC-seq data with Paired-Insertion Counting (PIC) and a model-based insertion rate estimator. Nature Methods 21.1 (2024): 32-36. doi:10.1038/s41592-023-02103-7.