Figure 4

TGIRT-seq of the Miltenyi miRXplore miRNA reference set using the NTT or NTC adapters and comparison of different methods for mitigating 5′- and 3′-end biases. TGIRT-seq libraries were prepared from the Miltenyi miRXplore miRNA reference set containing 962 equimolar human miRNAs (Supplementary Table S2 and Methods). Datasets for each method (three combined datasets for NTC, NTT, MTT, and NTT and a single dataset for NTT/6N) were used to plot both the empirical cumulative distribution function (ECDF) of the log₂ median-normalized counts for each miRNA ranked from least to most abundant (left panels), and the abundance-adjusted nucleotide frequencies at the 5′ end (positions +1 to +6) and 3′ end (positions −1 to −6) of the miRNA sequences in the dataset relative to those in the miRNA reference set (middle and right panels). Only uniquely mapped reads were counted. The numbers within the ECDF plots for each method indicate the root-mean-square error (RMSE) for over-represented miRNAs (top right), under-represented miRNAs (bottom left), and all miRNAs (top left). The curve plotted as a dashed line at the bottom of the ECDF plots indicates the distribution density of the 962 miRNAs in the dataset. (A) Miltenyi miRXplore reference set showing the ECDF plot layout (left panel) and the aggregate 5′- and 3′-nucleotide frequencies for all miRNAs in the Miltenyi miRXplore reference set assuming equimolar concentrations of the 962 miRNAs. (B–G) ECDF plots (left panels) and plots of the abundance-adjusted nucleotide frequencies at the 5′- and 3′- ends of miRNAs in TGIRT-seq datasets relative to those in the miRNA reference set (middle and right panels) for datasets obtained using (B) the NTC adapter; (C) the NTT adapter; (D) a modified NTT adapter mix in which the 3′ A overhang was replaced with a 3′ diaminopurine (denoted MTT); (E) a modified NTT adapter mix with an altered ratio of 3′ overhangs (A:C:G:T = 6.6:0.4:1:1; denoted NTTR); (F) the NTT adapter used in combination with an R1R adapter with six randomized nucleotides at its 5′ end (denoted NTT/6N); and (G) the NTT adapter after computational correction of 5′- and 3′-end biases (denoted NTTc).