Figure 3.

Outlier isomiRs can be identified by comparing the same isomiR sequenced on different platforms. (A) Heatmaps showing means of log2 transformed, z-scored values for tumor isomiR expression of patients’ tumor material analyzed on different sequencing platforms. A total of 115 samples sequenced on HiSeq were randomly chosen to equal the number of samples sequenced on GA. (B) After batch correction, variance is reduced. (C) Differences between expression values derived from different platforms are non-significant after batch correction. Association between log2 transformed expression values of the two different sequencing platforms for each isomiR. Student's t-test followed by FDR correction for multiple testing before and after batch correction (green and orange, respectively). (D andE) left panel: Means of log2 transformed isomiR expression values for patient tumor material sequenced on the GA (n = 131) or the HiSeq (n = 336) platform before and after batch correction. Outlier isomiRs show significantly different sequencing results on the different platforms (FDR corrected Student's t-test, padj < 0.05, log2 fold change difference >|0.5|) only before correction. Middle/right panels: Density plots for two exemplary isomiR expression values sequenced on either GA or HiSeq before and after batch correction.