Normalizing and denoising protein expression data from droplet-based single cell profiling

Abstract

Multimodal single-cell protein and transcriptomic profiling (e.g. CITE-seq) holds promise for comprehensive dissection of cellular heterogeneity, yet protein counts measured by oligo-conjugated-antibody can have substantial noise that masks biological variations. Here we integrated experiments and computational analysis to reveal two major noise sources: protein-specific noise from unbound antibodies and cell-specific noise captured by the shared variance of isotype controls and background protein counts. We provide an open source R package (dsb) to denoise and normalize CITE-seq data based on these findings. (https://cran.r-project.org/web/packages/dsb/index.html).