Skip to main content
. 2022 Apr 6;11:e76562. doi: 10.7554/eLife.76562

Figure 2. Clustering of 18 S rRNA modification profiles and correlation analysis from the mixture experiment and wild type rRNA.

(A) Hierarchical clustering of 18 S modification of profiles from pooled wild type, snR80ᐃ, snR83ᐃ, and snR87ᐃ RNA (500 reads in each experiment). Each row is a full-length molecule, each column is a modified nucleotide and the color represents modification probability, see scale. (B) Change in Spearman correlations of 18 S reads in the mixture experiment when compared to wild type. Stars represent significant changes when compared to wild-type correlation and significantly different from zero correlation (see Materials and methods). (C) Fraction of wild type, snR80ᐃ, snR83ᐃ, and snR87ᐃ profiles in each cluster of 18 S rRNA. (D) Table of snoRNAs knocked down with the corresponding expected knocked down modifications. (E) Hierarchical clustering of 18 S yeast rRNA modification profiles from wild-type yeast (1000 reads). (F) Wild-type Spearman correlation of 18 S wild-type reads. Stars represent significantly different to IVT correlations and significantly different from zero. correlation. (G) Crystal structure model of wild-type S. cerevisiae 18 S rRNA highlighting significant correlated positions. PDB: 4V88 (Ben-Shem et al., 2011).

Figure 2.

Figure 2—figure supplement 1. Heatmaps and percent modification change of snoRNA knockout and mixture experiments.

Figure 2—figure supplement 1.

(A) Heatmap of wild type, mixed sample, snR80ᐃ, snR83ᐃ, and snR87ᐃ, snR45ᐃ and snR4ᐃ modification profiles of 18 S (1,000 reads in each experiment). Each row represents a full length single read, each column represents a modified nucleotide and the scale represents the probability of being modified. (B) Mixed sample, snR80ᐃ, snR83ᐃ, and snR87ᐃ, snR45ᐃ and snR4ᐃ 18 S percent change in modification frequency when compared to wild type. Gray bars indicate the variance of wild type rRNA modification at each position and the black dotted lines represent the maximum variance found at any position. (C) Table of snoRNAs knocked down with the corresponding expected knocked down modifications. (D) Heatmap of wild type, mixed sample, snR80ᐃ, snR83ᐃ, and snR87ᐃ, snR45ᐃ and snR4ᐃ modification profiles of 25 S (1000 reads in each experiment). (E) Mixed sample, snR80ᐃ, snR83ᐃ, and snR87ᐃ, snR45ᐃ and snR4ᐃ 25 S percent change in modification frequency when compared to wild type.
Figure 2—figure supplement 2. Kmer distribution comparison between snoRNA knockout kmer distributions and the trained model kmer distributions.

Figure 2—figure supplement 2.

Each figure has the model’s canonical kmer distribution, the model’s modified kmer distribution and the corresponding snoRNA knockout kernel density estimate (KDE) of all events aligned to that position (see Materials and methods). The rows show kmers covering position 759 in 18 S from snR80ᐃ, position 776 in 25 S from snR80ᐃ, position 1290 in 18 S from snR83ᐃ, position 1415 in 18 S from snR83ᐃ, position 436 in 18 S from snR87ᐃ, position 436 in 18 S from snR87ᐃ, position 1773 in 18 S from snR45ᐃ and position 1280 in 18 S from snR4ᐃ.
Figure 2—figure supplement 3. Clustering and correlation analysis of snoRNA KO experiment modification profiles in yeast 18 S rRNA.

Figure 2—figure supplement 3.

(A/B) Change in Spearman correlations of 18 S reads in snR83 KO (A) and snR4 KO (B) when compared to wild type. Stars represent significant changes when compared to wild type correlation and significantly different from zero correlation (see Materials and methods). (C) Cryo-EM structure model of wild type S. cerevisiae 18 S rRNA highlighting significant, concerted positions. PDB: 4V88 (Ben-Shem et al., 2011). (D) Table of percent modification of sites of interest in wild type and snoRNA KO experiments.
Figure 2—figure supplement 4. Comparison of rRNA 2’O-methylation calling from other modification detection techniques and signalAlign modification detection.

Figure 2—figure supplement 4.

(A–B) Comparison between the range of modification percentages called via mass spectrometry (Taoka et al., 2016), HPLC (Yang et al., 2016), and two RiboMeth-seq approaches (Birkedal et al., 2015; Marchand et al., 2016) vs signalAlign modification percentages of wild-type yeast in 18 S (A) and 25 S (B). (C–D) Comparison between RiboMeth-seq modification percentages (Aquino et al., 2021) and signalAlign modification percentages for the Dbp3 knockout strain in 18 S (C) and 25 S (D) yeast rRNA. For the combination of several detection approaches, we calculated the minimum, maximum and mean modification percentage from the four papers. For all plots, error bars represent the minimum or maximum percent modification called and circles represent the mean modification percentage.
Figure 2—figure supplement 5. Yeast 25 S rRNA modification profile clustering and correlation analysis.

Figure 2—figure supplement 5.

(A) Hierarchical clustering of 25 S yeast rRNA modification profiles from wild-type yeast (1000 reads). Each row represents a full length single read, each column represents a modified nucleotide and the scale represents the probability of being modified. (B) Wild-type Spearman correlation of 25 S wild-type reads. Stars represent significantly different concerted positions compared to IVT and significantly different from zero correlation. (C) Crystal structure model of wild-type S. cerevisiae 25 S rRNA highlighting significant, concerted positions. PDB: 4V88 (Ben-Shem et al., 2011).