MiSeq, amplicons and clean-up

[AdaminMTL]

Hi All,

Say you want to amplify 75 small-ish (250-400 bp) loci from a population sample of, say, 15 individuals. So just over 350kb total you want sequenced by MiSeq with 2 x 250.

Question 1: Is there any particular reason adaptors and barcodes can't be ligated to the amplicons as they would be in genome sequencing?

Question 2: If they can be ligated as such, is there any reason to expect that they would (or would not) efficiently ligate to primer dimers in non-purified PCRs?

Basically, I'm asking if you could just amplify a bunch of loci for a sample of individuals and if the PCRs look good (i.e., no huge primer dimer problems), pool them and sequence them. The hope being that the differences in locus-to-locus (and individual-to-individual) amplicon concentration do not result in huge cluster differences and that the coverage overkill compensates anyway.

Don't worry, I haven't actually tried this... yet.

P.S. The motivation for this is what I perceive to be a great difficulty in getting accurate quantitation/normalization of 75 x 16 = 1200 samples without robotics. Also, I have seen firsthand some poor recovery with Ampure XP beads, not to mention quite a lot of sample-to-sample variation in concentration after PCR clean-up.


Sorry in advance for what may be blinding naivete.

- Adam

[JackieBadger]

Answer 1: No reason why not (for barcodes at least). This is what we do with MID tagged primers... just have non negligible start up cost
Answer 2: Not sure. In explicit regard to barcode tagged primers..they work well.

You can get relatively good quantification with a picogreen assay. No matter how hard you try to normalize sample concentrations, you will always have variance in converage.

[barrmur]

See if you can get access to a fluidigm access array. You might be able to prep all of the samples on a single chip on the access array in a much quicker time than by hand.

You could otherwise try running the 75 pcr runs, stick them on a gel to make sure they are ok, possibly assay using say a qubit, normilize and pool, then stick the lot into a standard truseq DNA prep at the end repair stage. 15 sample preps in total.

At the end of the day the 250bp PE run will give you huge coverage so you should be ok in terms of getting all targets covered even if some amplicons are over represented.

[JamieHeather]

I just wonder why, if you're amplifying anyway, why not include barcode and adapter sequences in the 5' of your primers and incorporate them all in one go?

[microgirl123]

The method in [this] (http://seqanswers.com/forums/showthread.php?t=32475) thread contains a two-step PCR method that avoids the ligation step.

[kmcarr]

I just wonder why, if you're amplifying anyway, why not include barcode and adapter sequences in the 5' of your primers and incorporate them all in one go?

For the experiment AdaminMTL described, 75 loci X 15 individuals, that would require 75 x 15 barcoded primers + 75 universal primers which works out to 1,200 primers.

This type of experiment is best performed using the two step (nested) PCR approach referenced by microgirl above. Then you only need 150 (75x2) target specific primers, plus 30 adapter/barcode primers. The adapter/barcode primers can be re-used for any set of properly designed target primers.

[JPC]

We've pooled PCR products for the MiSeq but had a terrible time getting them balanced, equimolar pooling was way off, it took a bit of 'trial and error' to get to an acceptable level

JPC

[Vinz]

If you optimize your PCR to give you reliable PCR products, you may skip normalization. With 15*75 amplicons you expect on average ~10000 reads (paired) per amplicon. That should be enough to get meaningful data for most samples and amplicons. Basically, the amplicons you would see on a gel will be represented by enough reads.
If the amplicons have a broad size distribution, the small ones will be overrepresented.
We use the 4 primer approach with 1 or 2 PCR steps without normalization which usually works fine.

[JamieHeather]

I've only done it a few times, but I've been running pools of two different length amplicons on the MiSeq. I quantify on Qubit, then use a Bioanalyzer to get the size and relative molarities of the two peaks, which I use to calculate a weighted average size and thus molarity. Seems to have worked pretty well so far.