Summary
A bioRxiv preprint reports that correcting ambient RNA before detecting doublets can leave more doublet-containing droplets in single-cell RNA-sequencing data. The authors recommend detecting doublets first, except when using CellBender.
A bioRxiv preprint reports that the order of two common quality-control steps can affect which problematic droplets are removed from single-cell RNA-sequencing data. In tests using engineered and multiplexed datasets, the authors found that doublet-detection tools generally performed better before ambient RNA correction, with CellBender as an exception. They recommend detecting doublets before correction unless using CellBender.
The study compared three ambient RNA removal methods—decontX, SoupX and CellBender—with up to six doublet-detection methods: scDblFinder, scds, bcds, cxds, Scrublet and DoubletDetection.
Why the processing order matters
Droplet-based single-cell RNA sequencing isolates cells in tiny droplets and measures RNA from each. Sometimes two cells enter the same droplet; these are called doublets. Their combined RNA can make a droplet appear to contain an unusual or hybrid cell type, so identifying doublets is an important quality-control step.
A separate complication is ambient RNA: free-floating RNA released by cells in suspension, likely during cell death, can be captured along with a cell. It can create unexpected gene-expression signals. Ambient RNA correction methods aim to remove this contamination, while doublet-detection tools try to identify droplets containing more than one cell.
The authors found that applying ambient RNA correction first—particularly with decontX—increased the proportion of doublet droplets retained compared with detecting doublets in uncorrected data. They also report that decontX and SoupX partly masked doublet signals: among doublets that remained undetected after correction, these methods removed proportionally more ambient RNA.
That matters because the two steps can influence one another. If correction changes the signals a detector uses to identify a doublet, the choice of processing order can affect which droplets pass quality control and remain for analysis.
The workflow recommendation and its scope
Based on their comparisons, the authors recommend running doublet detection before ambient RNA correction, except when using CellBender. The result is a workflow recommendation for the tools and datasets they evaluated, rather than a blanket rule covering every possible method or dataset.
The work is posted as a bioRxiv preprint. Its abstract describes the datasets and methods at a high level but does not give sample sizes or numerical performance estimates, so the reported findings here are qualitative.