To explore and visualise sources of variation in this dataset, we clustered a subset of 7,305 probes that were expressed in all 56 samples with a detection score >0.95 using principal components analysis (PCA) and hierarchical clustering methods. Figure 3 shows that PCA with three components separated the samples according to the five methods. The first two components in the PCA separated the PBMCs, B-cell CD19, B-cell CD20 from PAX and LCLs explaining 70% of the variance. The third component discriminated the PBMCs from B-cell CD19 and B-cell CD20 explaining 9.8% of the variance. Notably, B-cell CD19 and B-cell CD20 samples were clustered together.
