scDemultiplex: An iterative beta-binomial model-based method for accurate demultiplexing with hashtag oligos
Presenting author: Quanhu Sheng, Department of Biostatistics, Vanderbilt University Medical Center
Co-authored by:
- Li-Ching Huang, Department of Biostatistics, Vanderbilt University Medical Center
- Lindsey K. Stolze, Department of Biostatistics, Vanderbilt University Medical Center
- Hua-Chang Chen, Department of Biostatistics, Vanderbilt University Medical Center
- Alexander Gelbard, Department of Otolaryngology, Vanderbilt University Medical Center
- Yu Shyr, Department of Biostatistics, Vanderbilt University Medical Center
- Qi Liu, Department of Biostatistics, Vanderbilt University Medical Center
Abstract:
Single-cell sequencing has been widely used to characterize cellular heterogeneity. Sample multiplexing, where multiple samples are pooled together for single-cell experiments, attracts wide attention due to its benefits of increasing capacity, reducing costs, and minimizing batch effects. To analyze multiplexed data, the first crucial step is to demultiplex, the process of assigning cells to individual samples. Inaccurate demultiplexing will create false cell types and result in misleading characterization. We propose scDemultiplex, which models hashtag oligo (HTO) counts with beta-binomial distribution and uses an iterative strategy for further refinement. Compared with seven existing demultiplexing approaches, scDemultiplex achieved great performance in both high-quality