EmptyDropsMultiome discriminates real cells from background in single-cell multiomics assays

Area of Science:

• Single-cell biology
• Genomics
• Bioinformatics

Background:

• Multiomic single-nucleus technologies enable simultaneous measurement of chromatin accessibility (scATAC-seq) and gene expression (scRNA-seq).
• Accurate identification of true nuclei from background noise is crucial for reliable multiomic analysis.
• Existing computational tools may have limitations in sensitivity and cell-type bias.

Purpose of the Study:

• To develop and validate EmptyDropsMultiome, a novel computational approach for distinguishing true nuclei from empty droplets in multiomic single-cell data.
• To compare the performance of EmptyDropsMultiome against established methods like CellRanger-arc and EmptyDrops.
• To assess the statistical power, accuracy, and potential biases of different droplet calling strategies.

Main Methods:

• Development of the EmptyDropsMultiome algorithm for multiomic droplet analysis.
• Utilizing simulations to rigorously evaluate statistical power and accuracy.
• Application and comparison on real-world multiomic datasets.

Main Results:

• EmptyDropsMultiome demonstrates superior statistical power and accuracy compared to CellRanger-arc and EmptyDrops in simulations.
• Real-world data analysis reveals CellRanger-arc misses over 50% of nuclei identified by EmptyDropsMultiome.
• CellRanger-arc exhibits significant bias against certain cell types, with retrieval rates below 20% for some.

Conclusions:

• EmptyDropsMultiome provides a more sensitive and less biased method for droplet calling in multiomic single-cell experiments.
• The findings highlight the limitations of current tools and the need for improved background noise correction.
• This advancement facilitates more comprehensive and accurate single-cell multiomic data interpretation.