EmptyDropsMultiome Method Analysis

This repository contains the implementation and evaluation of the EmptyDropsMultiome method, as described in:

"EmptyDropsMultiome discriminates real cells from background in single-cell multiomics assays"
(Megas et al., 2024, Genome Biology)

EmptyDropsMultiome improves statistical power and accuracy in identifying real nuclei-containing droplets in droplet-based single-cell RNA + ATAC-seq data, outperforming CellRanger-arc.

Overview

This repository includes:

• Modified scripts to replicate simulations and benchmarking.
• Pipeline code to integrate emptydrops_multiome() into real-lab multiomic workflows.
• Downstream analysis using Seurat and Signac.
• Notebooks explaining method logic and code integration.
• Results from real and simulated datasets, including comparison to CellRanger-arc.

Key Components

EmptyDropsMultiome2023_final

Forked and extended version of the official repository from Megas et al. (2024).
Includes modified scripts to replicate and expand simulations and figures:

• simrun_multiomics.R
• simrun_multiomics_monocytes.R
• fig2.R
• fig2_monocytes.R

These scripts simulate data, evaluate recall/FDR across different configurations, and generate benchmark plots.

jupyter_nb/

Two Jupyter notebooks to help understand and visualize how the method works:

• granulocyte_analysis.ipynb: QC, visualization, and method usage example on granulocytes using EmptyDropsMultiome.
• multiomic_analysis.ipynb: Overview of the downstream analysis of RNA+ATAC data.

data/

Contains raw inputs, outputs from simulations and real data analyses, and reference files for downstream annotation and Seurat-based processing.

main/

This is the core working directory for the implementation pipeline:

• scripts/: Code that performs the two-stage analysis:
  1. QC and threshold selection for lower_rna, barhop_rna, lower_atac, and barhop_atac.
  2. Execution of emptydrops_multiome() using these thresholds on raw unfiltered data.
• outs/: Final results.

metadata/

Contains metadata and barcodes from CellRanger-arc runs.

results_monocytes/

Results of downstream Signac analysis (scRNA + scATAC) from monocyte data, including cluster-specific outputs and quality visualizations.

work/

All job codes used for running all the analysis. This includes logs and intermediate outputs for reproducibility and auditing.

Reference

Megas, S., Lorenzi, V., & Marioni, J. C. (2024). EmptyDropsMultiome discriminates real cells from background in single-cell multiomics assays. Genome Biology, 25(121).