Metatranscriptome recommendations: read length and depth

[TimK]

Hi all,

I have a couple of questions regarding metatranscriptomics of environmental samples.

1. What is the recommended read length?
I'm aware that for transcriptome data sets where a reference genome exists read length is often less important. However, given that metatranscriptomics include assembly of transcripts and subsequent mapping of reads back to the assembly I would assume that, as for metagenomics, metatranscriptome reads should be as long as possible.

2. Overlapping reads or not?
Am I correct that read overlapping, especially long/complete overlapping regions, are only useful for application cases where accuracy should be maximized, e.g. rRNA aplicon sequencing?

3. Sequencing depth:
I guess sequencing depth is always a tricky one and depends not only on the question but also on the environment (high/low diversity/complexity). But is there a better rule of thumb than "as much as possible"?

Thanks,
Tim

[Ali May]

I have some experience in comparative metatranscriptomics (e.g. disease vs healthy). From my point of view you have two options: HiSeq 2500 50 bp or 150x2 bp. The rule of thumb is that more depth never hurts in environmental sequencing, so if you have access to an instrument with even more yield, go for it. I try to answer your questions below:

1- If you want to do metatranscriptome assembly, needless to say, the read length becomes very important, as it is likely with very short reads to end up having many chimeric assemblies. But there is a trade-off between read length and sequencing depth, so an educated choice here is necessary. Also, depth as you mentioned is very closely linked to the environment of interest. You don't need very deep sequencing for a vaginal sample for instance, but the same doesn't go for oral samples, or soil etc.

2- Although I'm not very sure, I do not think that this is important.

3- In my opinion read length and sequencing depth should be considered together. If you have 40 different bacteria (2.5 -3 MB genomes) in your sample, I would suggest at least 30x. But, to be really honest, I am not sure if anyone can show you what the difference would be between 30x and 10x.

Well I didn't really say much than you already did, but I hope these help. On a different note: one thing to consider is to not do assembly... I've read and heard people saying that it's super important to assemble, because it lets you map many more reads to the transcriptomes. Fair enough... But I've never assembled and it worked just fine for me. especially if you have access to something like KEGG.