Fig. 4

Test for the robustness of the model. a Shown are effect of adding noise to data with 6 pairs used in Fig. 3b; b test on an independent set of 12 MCF7 RNA-seq datasets of GSE86316, with their first best-match MCF7, second best-match KLP-1 and the third candidate (not consistent for all 12 MCF7 samples); c test on an another independent set of 8 HCT116 RNA-seq datasets of GSE101966; and (d) effect of sequencing depth on prediction accuracy using subsampling (1 M, 2 M, 5 M, 10 M, 25 M and 50 M) method