Comparison of normalization approaches for gene expression studies completed with high-throughput sequencing - Fig 3

Plots of the first two PCs with the proportion of total variation for the CESC data to cluster primary factor of interest (keratin-low and keratin-high groups): (A, C) raw counts data, and (B, D) after adjusting the library size (UQ) and removing the effect of batch ID and estimated SVs by use of PCA through normalization. Each point is colored based on either the two keratin groups (biological factor) (A, B) or the14 levels of known batch ID (C, D).