What is the current recommended way to process/ analyze Ion Torrent Proton RNAseq data

Question

[cwwong13]

I read some older posts on this forum and others, where many people have complaints about the low alignment rate of Ion Torrent Proton RNAseq reads to the reference genome. However, when I tried to use 2-pass STAR alignment, I found that I can achieve on average 80-90% alignment rate.

I wonder if there is recommended way to process the Ion Torrent Proton RNAseq data? I found some literature mentioned using the software associated with Thermofisher (Torrent Suite), I wonder if that is the preferred way for processing these data? (If yes, I would like to know if there is an open-source alternative for the suite.

Lastly, I would like to know if I carry out the downstream analysis using DESeq2 or EdgeR is valid? Although that probably is quite repetitive to older questions, many of those questions are from date back in 2017. any updates on the recommended pipeline are greatly appreciated.

Thanks!

Answer 1

[Istvan Albert]

I don't think there is (or should be) a difference when analyzing Ion Torrent data vs Illumina data. Torrent suite is more of a software platform for managing the steps, rather than a substantial algorithmic innovation in my opinion.

In my hands edgerR seems more conservative than DeSeq2 - some caveats may apply. People use both all the time.

[cwwong13]

Would you mind explaining more why you think edgeR is more conservative and what are the caveats? While I have no preference on which of them, I do find that edgeR usually calls more features/ genes to be significant. (The difference is very tiny, I am not sure if that is due to some error on my analysis script though. Thus your explanation would be very helpful for my debug.)

[Istvan Albert]

For me, it is more of an empirical observation rather than a well-defined reasoning.

Kevin Blighe summarizes the main differences between the methods here:

How do I explain the difference between edgeR, LIMMA, DESeq etc. to experimental Biologist/non-bioinformatician

If I had to explain why I think edgeR is more conservative, feel that excluding the outliers produces makes the method less prone to errors.