scPDA: Denoising Protein Expression in Droplet-Based Single-Cell Data

Droplet-based profiling techniques such as CITE-seq measure the surface protein abundance of single cells, providing crucial information for cell-type identification. How-ever, these measurements are often significantly contaminated by technical noise, which lowers the efficiency of using the gating strategy to identify cell types. Current computational denoising methods have serious limitations, including a strong reliance on often-unavailable empty droplets or null controls, insufficient efficiency due to the ignoring of protein-protein interactions, and a heavy computational load. Here, we introduce scPDA, a new probabilistic model that employs a variational autoencoder to achieve high computational efficiency. scPDA completely eliminates the use of empty droplets, and it shares information across proteins to increase denoising efficiency. Compared to currently available methods, scPDA has removed noise much more thoroughly while preserving biological signals, and it has substantially improved the efficiency of gating-strategy-based cell-type identification, marking a clear advancement in the computational denoising of the protein modality.