Figure 3

Comparison of Illumina high-throughput sequencing and Illumina miRNA Bead Arrays v2 performance in thyroid tumor samples. Each dot represents the abundance of one mature miRNA in 1 of 20 samples in HTS (x-axis) and the fluorescent intensity of the corresponding spot on the microarray (y-axis) of all 20 samples. Expression values are based on normalized data, assuming perfect matching of microarray probe sequence and HTS sequence reads. The figure illustrates the saturation of microarray signals compared to the number of HTS reads and a lack of specificity of microarray probe hybridizations by the presence of fluorescent signal for miRNA targets without any HTS reads (x-axis position -∞). Because RPM normalization of low read counts generates visual artifact (column-like data structures between 0 and 5 RPM) we excluded all miRNAs with a very low abundance of one or two reads in individual samples. The correlation coefficient between the microarray and HTS data set is 0.67.