Automated denoising of CITE-seq data with ThresholdR.

Area of Science:

• Single-cell biology
• Immunology
• Bioinformatics

Background:

• Cellular indexing of transcriptomes and epitopes by sequencing (CITE-seq) integrates transcriptomic and cell surface protein data.
• CITE-seq data (antibody-derived tags, ADTs) can contain technical noise from ambient antibodies, non-specific binding, and titration issues.
• Accurate denoising is crucial for reliable interpretation of CITE-seq experiments.

Purpose of the Study:

• To introduce ThresholdR, an automated R-based tool for denoising ADT data from CITE-seq experiments.
• To systematically identify thresholds separating signal from noise for each antibody.
• To evaluate ThresholdR's performance and compare it with existing methods.

Main Methods:

• Development of an R package, ThresholdR, for automated ADT data denoising.
• Benchmarking ThresholdR against DSB and CellBender using diverse CITE-seq datasets and platforms.
• Assessment of ThresholdR's ability to reduce false negative rates.

Main Results:

• ThresholdR effectively identifies the signal-noise threshold for ADTs in CITE-seq data.
• ThresholdR outperforms DSB and CellBender by significantly reducing false negative rates.
• The denoising approach using ThresholdR enhances the accuracy of cell-type annotation.

Conclusions:

• ThresholdR provides a reliable and systematic method for denoising CITE-seq ADT data.
• Denoising CITE-seq data with ThresholdR improves downstream analyses, including cell-type identification.
• ThresholdR offers a valuable tool for researchers utilizing CITE-seq technology.