Thema der Dissertation:
Moving Beyond First-Order Approximations in Modeling and Analyzing RNA-seq Data Thema der Disputation:
Zero-inflated models for scRNA-seq data: are zeros biology or technical error?
Moving Beyond First-Order Approximations in Modeling and Analyzing RNA-seq Data Thema der Disputation:
Zero-inflated models for scRNA-seq data: are zeros biology or technical error?
Abstract: Single-cell RNA sequencing (scRNA-seq) data is dominated by zeros. This abundance of zeros was long considered as technical variability or missing data, and is most often modelled by Poisson and negative binomial distributions and their zero-inflated extensions. This talk introduces these count models and follows how three studies used them to revisit that view. Kim et al. (2020) show that a mixture of Poisson distributions always produces more zeros than a single Poisson with the same mean, so cell-type heterogeneity alone can appear as zero inflation. Choi et al. (2020) use Bayesian model selection to identify zero-inflated genes, and find that most lose their zero inflation once cell type is included in the model. Complicating the picture, Jiang et al. (2022) show that whether zero inflation is detected depends on the sequencing protocol, so it cannot be a direct readout of biological zeros. Together, these studies suggest that most zeros arise from sampling and biological heterogeneity, although the counts alone cannot fully separate biological from technical zeros.
Kim, T. H., Zhou, X., & Chen, M. (2020). Demystifying “drop-outs” in single-cell UMI data. Genome biology, 21(1), 196. Choi, K., Chen, Y., Skelly, D. A., & Churchill, G. A. (2020). Bayesian model selection reveals biological origins of zero inflation in single-cell transcriptomics. Genome biology, 21(1), 183 Jiang, R., Sun, T., Song, D., & Li, J. J. (2022). Statistics or biology: the zero-inflation controversy about scRNA-seq data. Genome biology, 23(1),31.
Zeit & Ort
12.10.2026 | 14:00
Seminarraum 2006
(Zuse Institut Berlin, Takustr. 7, 14195 Berlin)