Jonathan Goke from the Genome Institute of Singapore presented at the Nanopore Community Meeting in Singapore on “Identification of m6A RNA modifications at single molecule resolution using nanopore direct RNA-seq data.” They are working on methods to analyze transcriptomes and are interested in the epitranscriptome. Goke explained that m6A is one of the most frequent mRNA modifications that impacts RNA stability and translation. m6A deregulation is involved in cancer. Goke compared direct RNA sequencing to cDNA sequencing. m6A can be identified from direct RNA-seq data, but there is a challenge: “the multiple instance learning (MIL) problem” in which some RNA can be modified at one position but which ones are modified is not known. Goke’s team uses training data and evaluates the model. They first created a training model using kidney cells and applied to a human colon cell line. Reviewers suggested using Arabidopsis instead since the transcriptome has different kmer profiles. m6Anet accurately identified m6A from direct RNA-seq and generalized to new cell lines and species. The model does have to be re-trained for use with the new ONT chemistry: RNA004. m6Anet now supports RNA004, and it performs similarly to RNA002. Goke noted that they have an SG-NEx Project on AWS Open Data with tutorials for bambu, m6Anet, and xPore for transcriptome analyses.
