Evaluating and comparing transcriptome sequencing approaches - Rachel Goldfeder

Rachel Goldfeder of The Jackson Laboratory kicked off her presentation by stating that ‘direct RNA sequencing resolves challenges associated with cDNA sequencing’, citing the absence of reverse transcription and PCR, and the facility of direct RNA sequencing to enable the analysis of the native RNA molecule (allowing direct identification of base modifications), as key factors in this assertion.

Rachel gave an overview of her work in which she compared three different nanopore RNA sequencing approaches, namely direct cDNA, cDNA-PCR, and direct RNA, to the analysis of RNA from GM12878 (+SRIV spike-in control) and iPSC cells (+SRIV spike-in control).  The results of each method were compared for yield, read-length, and transcripts covered. On average, approximately 4 million, 13 million, and 2.5 million reads were generated for direct cDNA, cDNA-PCR, and direct RNA respectively. Average read lengths across all three approaches were similar, although, the cDNA approach offered very slightly longer read lengths of approximately 1 kb (compared with 800-900 bp for the other approaches). Using minimap2 to map the reads to a reference transcriptome (GRCh37 + SIRV transcriptome) provided highly concordant results across all approaches, with over 90% mapped reads.

Next, Rachel investigated whether the direct RNA approach, which involves sequencing from the 3’ end of the molecule only, leads to greater 3’ bias. The results of which showed similarly low levels of 3’ bias across all three techniques.

Direct RNA sequencing was found to cover fewer transcripts than the cDNA approaches due to the lower number of reads generated. Examining the transcripts in more detail it was evident that the vast majority were shared between all approaches and any discrepancies may be a result of different coverage levels. Furthermore, all three methods showed high levels of concordance with regard to transcript expression levels. The number of reads per transcript were also highly correlated across techniques and replicates for both cell types tested.

Presenting the SRIV spike-in data, Rachael commented that the ‘quantification matched well with expected values’ across all RNA sequencing techniques; however, the direct RNA sequencing approach showed the most precise quantification.

The team at Jackson Laboratory then sought to evaluate the identification of different isoforms using the SIRV sequence data together with their novel FLAIR analysis algorithm, which was presented in more detail by Alison Tang in yesterday’s Transcriptomics breakout session. The sensitivity for exonic bases was shown to be high in all libraries, with direct RNA libraries displaying the highest levels of precision.

Rachel dedicated the final part of her talk to the use of direct RNA sequencing for modified base detection. She first highlighted the importance of RNA modifications in regulating gene expression, altering RNA function, and RNA stability. Through these actions, RNA modifications have been shown to modulate cell survival, differentiation, and migration – with implications for disease development and drug resistance. Rachel stated that one challenge researches face when analysing RNA modifications is the ‘lack of proper positive and negative control sequences’. Rachel presented the results of a collaboration with Dr. Kin Fai Au and Dr. Yunhao Wang at The Ohio State University to address this challenge. The team generated and sequenced negative (unmodified) and positive (modified) controls. The direct RNA sequencing reads were then aligned to a reference using minimap2. The raw nanopore signal was analysed using nanopolish, revealing distinct separation of signal between modified and unmodified bases and, thereby confirming that computational models and methods can be built to identify modified bases. According to Rachel ‘the future of modification detection is bright’.

In closing her presentation, Rachel presented a useful overview table of her RNA sequencing results, highlighting the benefits of each of the three Oxford Nanopore RNA sequencing approaches for different applications — with the assertion that the best approach ‘depends on your needs and use case’.