Journal: Nature biotechnology

Article Title: Comprehensive multi-center assessment of accuracy, reproducibility and bias of small RNA-seq methods for quantitative miRNA profiling

doi: 10.1038/nbt.4183

Figure Lengend Snippet: (a) Boxplots show the observed ratio (y-axis; log2-scale) vs. expected ratio (x-axis) for miRNAs present in each of the SynthA and SynthB synthetic RNA subpools. Observed ratios for each miRNA were calculated as mean CPM of SynthA/mean CPM of SynthB across technical replicates for each lab + library prep method. Boxes show the median + IQR; upper/lower whiskers indicate the smallest/largest observation less than or equal to 1 st /3 rd quartile -/+ 1.5 * IQR; outliers are calculated as being < 1 st quartile – 1.5 * IQR or > 3 rd quartile + 1.5 * IQR. Mean CPM ratios were calculated from n=4 SynthA and n=4 SynthB technical replicate libraries for each lab and library preparation method shown, with the exception of TruSeq Lab8 SynthB (n=3). Those miRNAs with a mean CPM of 0 in SynthA or SynthB are not plotted. The numbers of miRNA not plotted are as follows: Truseq Labs 1,2,3,4,5,6,8: 1; Lab9: 0; CleanTag Lab5: 0; NEBNext Labs 1,3,5,9: 0; Lab4: 1; Lab2:3; 4N_NEXTflex Lab7: 1; other 4N: 0. The number of sequences represented in each boxplot is provided in (b) Heatmaps show the pairwise, squared Spearman rank correlation coefficients from sequencing the SynthA and SynthB pools. Pairwise correlation coefficients were calculated based on the mean CPM across technical replicates for SynthA samples (left), SynthB samples (middle) and the ratio of SynthA : SynthB (right). The mean CPM value for each ratiometric pool sequence was calculated from n=4 technical replicate libraries per lab, library preparation method and pool. Mean CPM values for n=290 ratiometric pool RNAs were used for calculating each pairwise correlation coefficient. Hierarchical clustering for rows and columns is the same for all heatmaps, and is based on the average pairwise Euclidean distances calculated from the SynthA CPM and SynthB CPM correlation matrices. Column labels indicate the lab ID and library prep method; row labels indicate only lab ID, but are presented in the same order (top to bottom) as columns (left to right).

Article Snippet: Single-end libraries were sequenced using the Illumina HiSeq 2500 (Lab8 and Lab9 for all the protocols and Lab5 for TruSeq, CleanTag, 4N-Xu and 4N_A), Illumina HiSeq 4000 (Lab4 for all the protocols and Lab1 for TruSeq, 4N_D and equimolar NEBNext), Illumina NextSeq 500 (Lab2, Lab3 and Lab6 for all the protocols, Lab5 for NEBNext and in-house 4N_B and Lab1 for ratiometric and plasma NEBNext) or Ion Torrent (Lab7) platforms (see which also includes information on miRNA editing libraries).

Techniques: Sequencing