Examination of PCR biases in 16S experiments.

[ScottC]

Hi Everyone,

I was just thinking about things that might affect the results of 16S amplification and sequencing in terms of lab-based biases. Do you take any of these into account when designing and executing experiments? Have you ever measured their effects?


1. Real base balance in oligo synthesis where an ambiguity is specified: Bases don't have equal coupling efficiencies, so 'N' in the oligo order doesn't necessarily equate to 25%A, G, C, T at that position across the oligos in the pool).

2. Primer synthesis termination and deletion products: As little as 30% of synthesised oligo is full-length and is the expected sequence. The rest is early termination products and deletion products.

3. Amplification efficiency/exponential amplification/reaction kinetics - It seems that there's a much higher likelihood of chimera formation, template amplification suppression and introduction of other biases when the reaction is 'inefficient' for whatever reason (such as 'tired' PCRs where reagents are running low in the reaction due to excessive cycling).

4. Non-specific amplification.

5. Polymerase choice and enzyme error profiles.

... anything else you consider to be a problem? What else do you evaluate, measure and/or control?

Cheers,

Scott.

[cliffbeall]

One thing you might look out for is contamination from previously run samples, etc. We have started pooling samples in such a way that we can detect barcodes that should not be there, and looking for all barcodes in every run.

Sequencing multiple regions of the 16S gene can be useful to confirm that there are not discrepancies. Or compare to whole metagenome sequences (for at least a few samples)?

Our experience is that mock communities are not great as a control, that is also reflected in some published data I have seen. Not sure what the explanation is.

[ScottC]

Another thing to add:

For point 4 in my original post: Many oligo synthesis companies do not guarantee that their oligos are free of DNA and other oligo contamination. We were quoted 0.5% 'crosstalk' in oligo synthesis (I'm not sure exactly what that means).

[Palecomic]

Biases are endless, and boil down to - keep everything the same.

Lots have been covered in the TRFLP and DGGE literature, which were also PCR based, and biases that are detectable using those methods are going to be even more problematic with sequencing.

Batch and manufacturer of Taq have shown biases, you can pool aliquots for a single study if you're worried.

We get HPLC purified primers

We perform the PCRs in quadruplicate and pool the products before sequencing (this was demonstrated to be useful years ago by Colleen Cavanaugh's group, but may or may not be necessary, I know some people are revisiting).

We use a high fidelity Taq, though I've seen something recently that proof-reading can lead to greater chimera formation.

On the sequencing side, a software update on your 454 definitely causes problems (you get different reads out the other end for the same run and therefore different OTUs)

Contamination is definitely a big problem, particularly with low biomass samples. We sequence our PCR negatives and include kit controls (the sampling vessel and medium has a DNA extraction done, gets PCR amplified and sequenced) - you almost always find bugs in kits. Delftia, Sphingomonas, Herbaspirillum are common ones.

We recently had a big contamination problem and tracked it down to a PCR additive contaminated at the manufacture and variable within the lot. Bioron, Molzym and Mobio do produce reagents that are guaranteed bug DNA free, but not enough to cover every step and in which case, why not put up with it and identify what it is?

Interested to hear about the mock community Cliff, I was going to start assembling one shortly as a per run control - what do you use as the template, genomic DNA or clones or other amplicons? We made some very simple ones here that worked out OK http://www.plosone.org/article/info:...l.pone.0032543 but I want something a bit more involved.

[ScottC]

1. Real base balance in oligo synthesis where an ambiguity is specified: Bases don't have equal coupling efficiencies, so 'N' in the oligo order doesn't necessarily equate to 25%A, G, C, T at that position across the oligos in the pool).

So it turns out that, when ordering 'standard' oligos, IDT's products have a base balance whereby 'N' really means A:20%, C:24%, G:26%, T:30%.

I wonder if this bias would be noticeable in datasets depending on the primers used.

[Genohub]

randomized adapters

I highly recommend not purifying by HPLC if you are interested in retaining an even representation of randomized bases (N), e.g., 25%A, 25% T, 25%G, 25%C. Purification can skew these populations, removing the highest and lowest MW species. That being said, if the bases are on the 5'end, and you are synthesizing a long oligo, then not purifying can also be problematic as a certain % of your oligo species will be truncated. If they are closer to your 3' end then its less of a problem because oligos are synthesized 3' to 5'.

[ScottC]

Hi,

IDT actually recommended against any type of purification, because they all skew the composition in different ways. They actually recommended paying for a hand-mixed reagent when requesting ambiguous bases. They mix them in the right ratios to achieve as close as they can to even representation of oligos.