Fig. 4
From: A general and flexible method for signal extraction from single-cell RNA-seq data
Average silhouette width in real data sets. a V1 data set; b S1/CA1 data set; c glioblastoma data set; d mESC data set. Silhouette widths were computed in the low-dimensional space, using the groupings provided by the authors of the original publications: unsupervised clustering procedure (a, b), observed characteristics of the samples, such as patient (c) and culture condition (d). PCA and ZIFA were applied after normalization: unnormalized (RAW), total count (TC), trimmed mean of M values (TMM), and full-quantile (FQ). For the mESC data set (d), we fitted the ZINB-WaVE model with batch as a sample-level covariate (ZINB-Batch) in addition to the default model with only a sample-level intercept (see Fig. 5)
