Sequencing depth

[rEDI]

I am sequencing samples from animal gut using illumina MiSeq/HiSeq, targeting the v4 region of the 16S

I get 22m raw reads back from MiSeq (after removal of PhiX etc), for a total of 24 samples.

I am getting an average of ~400,000 classified reads per sample after filter steps have been performed in Mothur.

I then rarefy to the sample with lowest no of reads (approx 150k seqs).

The rarefaction curves plateau after approx 60k reads or so.

Do you think the sequencing is too deep, and if so is this a big problem?

Thank you

[fanli]

"Too deep" sequencing generally isn't considered a problem. In any case, you're rarefying your data to 150k sequences which is fine. For comparison, we target 100k/samples for our V4 sequencing.

[rEDI]

Thank you Fanli for your answer

In another study, I rarefied to 300k seqs/sample. Is this also acceptable?

Thank you

[rEDI]

Hi Fanli

Do you think that rarefying my data to 300k seqs/sample is also ok, or too deep?

In one run I obtained slightly more reads passing filter, and so I rarefied to that sample containing the the lowest no of reads (apprx. 300k).

Some studies use much lower no. of reads, whilst others use numbers comparable to mine. I suppose it also depends on the number of samples sequenced (i.e. 24 samples will produce more seqs passing filter than 96 samples)?

Please let me know what you think.

Kind regards

[fanli]

No, unless I'm missing something rarefying to a higher depth is not a problem.

I'm curious though, why are you doing so much sequencing? Is your environment extremely high biomass?

[rEDI]

No, the reads usually plateau after 60k or so per sample.

I am analyzing fish gut samples.

After v4 sequencing on MiSeq, I get 22m reads for 24 samples. My feeling is that if I sequenced more samples I would have less reads per sample and hence would not be analyzing such a high number?

After filtering and QC, I usually get about 3-500k seqs/sample.

Therefore rarefying to ~300k seqs/sample for analysis.

Hope that explains it?

[fanli]

Yeah, if I were you I'd multiplex a lot more samples per run. We generally do anywhere between 70-120 samples.

[rEDI]

Ok thanks Fanli, will do that for the next sample set

Even still, is rarefying to this higher depth with my current number of samples ok?

[GenoMax]

Ultra deep sequencing data can cause problems with assemblies. See this paper.

[fanli]

@GenoMax, yeah that fact is still a bit hard for me to internalize. MORE is always better right?? (Murica, sorry).

@rEDI, if you are going to analyze all your samples together, you should rarefy all of them to the same depth. Ideally, you want positive and negative controls as well to quantify robustness and contamination. Although if your plateau is ~60k reads, you probably don't have to worry as much about the kitome.

[rEDI]

Thanks Genomax and Fanli

I recently read a paper that sequenced 96 samples of fish gut on MiSeq, with V4 primer set, the exact same as mine.

That study obtained ~3m reads in total after QC. Therefore approx 32k reads/sample.

I only sequenced 24, therefore obtaining the higher number of reads per sample.

Looking at my reads, had I sequenced 96 samples, I would have received comparable numbers to the other study (~approx. 25k/sample).

The read assembly is fine, just high numbers of reads per sample.

Am I right in saying this is due to low number of samples multiplexed (24), and that I can just rarefy to the lowest number of QC'd and assigned reads (150k in one experiment, 300k in another)??

What do you think?

Thank you

[rEDI]

If anyone could offer some insight into the above comparison between seq depths that would be great

Thank you so much

[thermophile]

You need to rarify all of the samples that your are analyzing together to the same depth. Even if you are sequencing close to saturation, you can never know if a zero is truly absent from your community or simply not captured.

On your broader question I think 100k is pretty excessive for microbiome and you're likely sequencing a lot of PCR duplicates. (remember a decade ago we were looking at communities with 50-100 clones and in general finding the same broad patterns. More samples will likely give you more insight to your question than more sequences per sample)

[rEDI]

Thanks thermophile - really appreciate it

If that is the case, would sequencing PCR duplicates impact upon the downstream analysis, or preserve pretty much the same trends as rarefying to lower depth?

[thermophile]

sequencing pcr dups isn't good because there is going to be a strong phylogenetic signal in that bias-the "universal" primers aren't perfect matches to all bacteria, so groups that match better will likely amplify quicker which will inflate their abundance. This is one of the big caveats to microbiome sequencing. Excessive sequencing depth doesn't make the problem worse but it can lead to a false sense of strength of your findings and it's expensive.