Fig. 2
From: The Poisson distribution model fits UMI-based single-cell RNA-sequencing data
The Q-Q envelope plots for different Poisson parameters (a–f) for the different degrees of batch variation. The plots indicate that the IPD statistical framework fits the data well, where most deviations were explained by an inappropriate choice of the number of latent vectors \(L\). Plates 1, 3, 5A, and 5B are biological replicates of the same clonal cell line. Plate 6A is from a different clonal cell line. Plate 8 has two cells per library. The doublets in Plate 8 required a larger \(L=15\) (dark green) than the default \(L=10\) (light green). The mixture cell lines from Plates 5A and 6A is better modeled by \(L=20\) (red) than \(L=10\) (orange)