Metagenomic amplicon sequences-to-sample misidentification description and solution

[LejzerowiczF]

Hi All,

I cannot resist asking the community about its feeling/experience on the underrated issue we've been studying (open access: http://nar.oxfordjournals.org/conten.../18/nar.gkv107).
It's surprising that loads of cross-contaminations created by the switching of tags/indices sequences appended to amplification primers and used for sample demultiplexing aren't addressed much... Did you noticed this in your metagenomic amplicon datasets?

Cheers to all!

[bunce]

Dear Franck,
A really fascinating paper - thanks for sharing - which together with this recent paper ( http://onlinelibrary.wiley.com/doi/1...12402/abstract) really show up the issue. Clearly people need to think hard about the workflows.

you asked for feeling/experience so I thought I would chip in a few (well, five) thoughts.

(1) completely agree you need to get tags onto samples in the first round of PCR. To not do this is molecular biology madness.

(2) Your 2nd round of PCR (8 cycles is still a long-primer PCR) - less cycles, granted - but still subject to the same bias and a 1-step long-primer approach.

(3) So I guess I only disagree with your recommendation VII in the conclusion (which you don’t present data on)…..there are issues (primarily efficiency, particularly with low-template samples) with long-primer approaches to library generation - but this tag jumping does not happen if you do go down this 1-step path. You might get intra-sample chimeras but not inter-sample ones (the latter of which are a much larger issue). We routinely spike-in untagged primers into these reactions - this may help to alleviate some of the efficiency drops. We monitor all steps using qPCR and don’t re-use the same forward/reverse index to avoid contamination as we are doing sensitive work (aDNA etc). I guess there is a trade off in efficiency with long-primer approach v's mis-tagging/contamination risks with the 2-step approach…. in my mind it is not a question of what is ‘better’ they are different.

(4) You cite the Berry paper as the reason for your recommendation #VII. Which is an interesting look at 1-step v’s 2-step approaches But it is a single primer set, has no spike-in of untagged primers, no qPCR to look at template input/efficiency and the differences encountered are not huge…. (fig 2a). I like this paper, but it is not ‘definitive’…. and in fact some of the ‘extra' diversity using a 2-step my have in fact been generated by the mis-tagging that you suggest.

(5) In our lab we use either (a) single-step long-primer fusions on ‘good’ samples… and (b) primer+index PCRs which we ligate adapters on in a PCR-free ligation. Through this set-up we are (desperately) trying to avoid a 2nd round PCR. 2nd Round PCRs are (as you paper elegantly points out) the source of mis-stagging and are also a contamination risk.

So, thanks again for the paper - the LSD solution to the mis-tagging is neat…. but a carefully implemented 1-step approach can side-step many of the issues with this as well.

Cheers, Mike