Long read sequencing reveals changes in measured isoform structure and abundance due to RNA processing

Abstract

Oxford Nanopore Technology enables direct RNA sequencing without reverse transcription (RT) and cDNA sequencing without PCR amplification, allowing investigation of how RNA preparation protocols bias sequencing data. PCR amplification is known to introduce biases related to expression levels, transcript lengths, and GC content. However, biases introduced by RT have been described more recently and their quantification remains more elusive. Using seven cell line datasets from SG-NEx, we compared nanopore sequencing of direct RNA, direct cDNA, and PCR-amplified cDNA. We identified hundreds of differentially expressed genes and isoforms between protocols, indicating significant changes in observed expression. Differentially expressed isoforms displayed expected amplification biases, and further we observed decreased PCR counts associated with higher m6A methylation levels on native RNA. We also found differential transcript usage in hundreds of genes across cell lines, showing substantial changes in measured isoform proportions between protocols. Analysis of structural differences between isoforms that switch dominance revealed that while some events involve simple RNA fragmentation, most involve complex structural differences including novel junctions (e.g. exon skipping, intron retention). Compared to non-DTU genes, inserted introns (exitrons) were significantly more common after RT and amplification. Furthermore, visualisation of raw coverage data proved crucial for qualitative validation of experimental findings.