Figure 1.
Example of the qualitative and quantitative results from Agilent’s Bioanalyzer 2100.
In panels (A) and (B) we show peaks produced from electropherograms (top left) that depict the size distribution of RNA fragments, the corresponding gel-like image of RNA fragments (top right), and metrics of RNA concentration and integrity (r26S/18S and RIN). We show two representative samples with (A) high-quality RNA in terms of high yield and minimal degradation (r26S/18S ≥1, RIN ≥8), and (B) low-quality RNA in terms of modest yield and high degradation (r26S/18S <1, RIN <5). Absence of clear peaks at 26S and 18S and an abundance of short fragments clustering on the left of the electropherogram in panel (B) are the hallmarks of severe RNA degradation. In electropherograms, ribosomal 26S (large) and 18S (small) subunits are shown in green and pink, respectively. Concentrations of 26S and 18S were calculated by taking the area above the green and pink straight lines at the base of the 26S and 18S peaks, respectively.
Figure 2.
Schematic representation of the method used to assemble Illumina reads into contigs, and contigs into scaffolds.
All reads were initially assembled into contigs using the de Bruijn graph method without using information about paired-end reads (shown by blue dashed lines). A contig’s sequence was resolved at every base. Contigs were then assembled into longer scaffolds by connecting contigs that contained paired-end reads assembled into separate contigs. Assembling scaffolds in this way allowed us to create longer sequences of known length, but sometimes there were gaps of unknown sequence. These gaps were constrained to represent <5% of total sequence length.
Table 1.
Effects of tissue type and age on metrics of RNA quality and sequencing.
Figure 3.
Variation among tissue types in RNA quality and transcriptome size.
We observed differences among tissue types for (A) total RNA mass (µg) isolated, (B) RIN, (C) sequencing depth and (D) number of scaffolds. For each tissue, we show the mean +1 SE and sample size at the base of columns. A posteriori pairwise contrasts among means corrected for multiple comparisons are shown in Supplemental Tables 3 and 5.
Table 2.
Statistical significance of explanatory variables in the best-fitting models for the data set with OD ratios and without OD ratios.
Figure 4.
Factors that significantly predicted the number of large scaffolds.
Among our measures of RNA quality, (A) RNA integrity number (RIN) and (B) OD 260/230 ratio were the strongest predictors of the number of scaffolds ≥1000 bp. (C) Sequencing platform also had a strong effect on number of large scaffolds (P<0.001, Table 2; numbers at the base of bars show sample size), and (D) mass of RNA sequenced had a weak but detectable effect (see Table 2). Note, for most samples we used 20, 30 or 40 µg of total RNA for sequencing, but a few samples used intermediate or lower amounts.