Chenchen Zhu from Stanford University presented at London Calling 2022 on the “Systematic assessment of long-read RNA-Seq datasets and its application in transcriptome analysis.” They are interested in understanding RNA isoforms using long-read sequencing. Differences in start sites and isoforms can lead to disease. The experimental workflow that Zhu shared was using reverse transcription and PCR before library prep. The average length is abous 2 kb. Zhu also shared a transcriptome annotation using full-length transcript sequencing. It is important to remember that Zhu’s workflow uses and oligo dT primer for reverse transcription. Zhu did share challenges in transcript identification: not all reads are full-length, truncated transcripts due to reverse transcription and RNA degradation, and sequencing errors. Sometimes exons don’t align, and Zhu explained that “microexons are often missed in long-read RNA-seq.” These errors can cause frameshifts. Zhu found that shorter exons smaller than 60 bp seem to be more prone to these errors. Zhu implemented a “rescue” by performing realignment of suspicious misaligned regions. Zhu concluded that rescuing microexons leads to better annotation. This work emphasized the importance of filtering and performing careful analysis of transcript counts to remove false positives. Zhu said that “most of the noise is actually random.” This is thought-provoking, and I am curious why!
