Skip to main content
. 2018 Jul 24;5:180136. doi: 10.1038/sdata.2018.136

Figure 1. Study design.

Figure 1

Preprocessing of raw data from ten independent datasets was done for normalization, background correction and probe-to-gene mapping. The fRMA-normalized data were corrected for batch effect using ComBat method and filtered for genes with low variance across samples. Validation of our dataset was done with PCA analyses and similarity measurement using RNA-Seq-profiled samples. Statistical R packages used to develop this dataset are stated.