Figure 8

Representation of the Miltenyi miRXplore miRNA reference set in datasets obtained by TGIRT-seq with the NTT adapter before and after computational correction compared to representation of the same miRNAs in datasets obtained using 4N protocols. (A) miRNA representation for TGIRT-seq NTT versus 4N. Log₁₀CPM values for each miRNA in combined TGIRT-seq NTT datasets (n = 3) are plotted against those in combined datasets for 4N protocols (n = 24; Gilardez et al.³⁶). Each point represents one miRNA. (B) The same comparison as (A) after computational correction of the TGIRT-seq NTT dataset using the random forest regression model (Fig. 5C,D). In (B), miRNAs are color-coded by their lengths (scale to the right). The purple dotted lines delineate 95% confidence intervals (2 standard deviations from the mean) of the miRNAs for 4N (vertical dotted lines) or NTT (horizontal dotted lines). The box formed by the intersections of the dotted lines encompasses 892 miRNAs that lie within these confidence intervals and were used for comparison with over- and under-represented miRNAs in Fig. 9. The expected CPM values (CPM = 1,039.5 for each of the 962 equimolar miRNAs) are indicated by horizontal and vertical orange lines for TGIRT-seq and the 4N protocols, respectively. The diagonal red line indicates cases where the CPM values from NTT are equal to those for 4N protocols. (C) Correlation between miRNA abundances and miRNA length. Two-dimensional kernel density estimation of the distribution for miRNA abundances and lengths (n = 962) is shown. The linear regression, with the equation: log₁₀CPM = 0.09 (miRNA size) +0.9, is plotted as a red line, and miRNAs with length <21 or >23 nt are indicated as white crosses. The coefficient of determinant (R²) is indicated in the plot. The color scale indicates the numbers of miRNAs not shown as crosses.