Fig. 2

Data overview in the space defined by principal and independent components. Data variability captured by the first components of PCA (a) and two selected components of ICA (b) in gene expression data. Independent components were selected based on the predictive power of their weights for patient gender (RIC3) and sample type (RIC5). MiRNA data showed even higher discrepancy comparing miRNA-seq and qPCR results by PCA (c). However, in the space of independent components (MIC1 and MIC9), the samples studied by miRNA-seq and qPCR overlap (d)