Common Analysis of Direct RNA SequencinG CUrrently Leads to Misidentification of 5-Methylcytosine Modifications at GCU Motifs

Abstract

RNA modifications, such as méthylation, can be detected with Oxford Nanopore Technologies direct RNA sequencing. One commonly used tool for detecting 5-methylcytosine (m⁵C) modifications is Tombo, which uses an "Alternative Model" to detect putative modifications from a single sample. We examined direct RNA sequencing data from diverse taxa including virus, bacteria, fungi, and animals. The algorithm consistently identified a 5-methylcytosine at the central position of a GCU motif. However, it also identified a 5-methylcytosine in the same motif in fully unmodified in vitro transcribed RNA, suggesting that this a frequent false prediction. In the absence of further validation, several published predictions of 5-methylcytosine in human coronavirus and human cerebral organoid RNA in a GCU context should be reconsidered.

Figure 1.

ONT direct RNA sequencing and GCU motif detection. A Schematic showing how RNA molecules are directly sequenced with ONT, followed by basecalling of the signal data produced by changes in ionic current, mapping to a reference, and detection of modified bases. Adapted from “Nanopore Sequencing”, by BioRender.com (2022). Retrieved from https://app.biorender.com/biorender-templates. B MEME Suite motifs overrepresented in the 10-nucleotide sequences surrounding the top 1000 putative modifications detected by the Tombo m⁵C “Alternative Model” for B. malayi, D. ananassae, C. albicans, and E. coli.

Figure 2.

Methylated fractions predicted by Tombo “Alternative Model” for m⁵C. A Density plots of the methylated fraction at all 3-mers containing a central cytosine in native and IVT viral RNA. Cytosine positions were filtered for depth > 10 and methylated fraction > 0 in both IVT and native samples. Histograms available in supplemental information. B Boxplot showing distributions of the methylated fractions detected by Tombo m⁵C “Alternative Model.” The methylated fraction was extracted for cytosines in animal, fungal, and bacterial RNA with depth > 100 and cytosines in native and IVT viral RNA with sequencing depth > 10. Methylated fractions were grouped based on non-GCU and GCU sequence context. Statistical significance based on p-value from two-tailed Z-test and Cohen’s d effect size.