Fig. 7
From: Quantile normalization of single-cell RNA-seq read counts without unique molecular identifiers
GLM-PCA dimension reduction applied to quasi-UMI counts reduces batch effects in the non-UMI Patel glioblastoma dataset. GLM-PCA was designed for UMI counts. In the top four panels, PCA was applied to normalized counts, and all normalizations failed to merge the two batches of tumor MGH26. In the bottom two panels, GLM-PCA was applied directly to census counts and QUMI counts (Poisson-lognormal with shape 2.0)