RNAseq by nanopore sequencing versus Illumina

[Fitzcarraldo]

Can I please ask for advice on how best to compare, on paper first of all, setting up an RNAseq experiment and the relative pros and cons of Illumina short read sequencing versus Oxford nanopore long read sequencing.

From the literature there appears far fewer reads for the latter but these reads are much longer and provide information on splice variants.

Previously we have done RNAseq by Illumina kits and we performed a rRNA depletion on human total RNA (5 ug) then a TruSeq stranded total RNA kit prep then onto a NovaSeq6000 with 4 biological replicates and acquired ~40 M reads for each. What would this depth possibly look like in an ONT experimental setup is my question?

Using ONT approaches and attempting isoform detection in mRNA transcripts then either PCR-cDNA Sequencing Kit (SQK-PCS111) or Direct cDNA Sequencing Kit (SQK-DCS109) seems to be the way to go for RNAseq. The choice seems to comes down to how much starting RNA you have and if you want to avoid PCR of the cDNA and what your output needs are. PCR-cDNA sequencing (SQK-PCS111) gives the highest output of reads and requires 4 ng of PolyA mRNA and has a newer chemistry which reduces intron priming during reverse transcription. Direct cDNA sequencing (SQK-DCS109) avoids PCR amplification of cDNA but requires 100 ng PolyA+ RNA and doesn't have the newer generation chemistry of SQK-PCS111.

Many thanks in advance for any advice or comments or kindly pointing me in the right direction in the literature!

[gringer]

What would this depth possibly look like in an ONT experimental setup is my question?

You haven't given enough information about the NovaSeq runs. What chip? What read length? Paired end?

Are you paying the same amount for ONT as for NovaSeq? If so, you'd likely be using PromethION flow cells, in which case the ONT per-base cost is similar to a NovaSeq S2 chip. ONT reads for cDNA average around 600bp for well-prepared fresh samples, so assuming this is being compared to 75bp single-end Illumina runs, you would expect about 5M reads per sample with a similar cost.

Something else that is worth bearing in mind is that the ONT kits don't produce any intronic reads. These can occupy a substantial proportion of Illumina sequencing runs, and have an impact on transcript count estimation. From a base-level transcript coverage perspective, ONT reads have a much flatter profile, which also helps for quantification and isoform discovery.

When I do sequencing on MinION flow cells ($1000 USD / run), I aim for at least 1M reads per sample. This is usually achievable without much effort when multiplexing up to 6 samples per run with the PCR-cDNA barcoding kit. I wouldn't recommend single-sample kits (e.g. PCS111), because I haven't found much benefit for downstream analysis for over 1M reads per sample. ONT claims that their newer Kit14 kits should have even better reliability (thanks to an increased sensitivity adapter and the use of UMIs), so it might be possible to do 12 samples from a single MinION flow cell with >1M reads per sample.