Hi,
I have been pondering this question for a while and was wondering what people's thoughts were? There seem to be multiple different methods of dealing with OTUs found in extraction kit and PCR controls, many of which have some pretty obvious problems. Obviously we can't just ignore that contamination may be present in lung, or indeed any low biomass, microbiota samples.
One fairly common way of dealing with this problem is just to remove all OTUs found in the controls from the samples. This seems a little drastic to me, particularly as bacteria such as Burkholderia and Corynebacterium are common contaminants but also contain important lung associated species.
Anther common method is just to report the types of bacteria found in controls and samples and do a statistical comparison to prove that the bacterial communities found are different. This is probably worthwhile doing just to prove that your results aren't totally due to contamination. However, if you are wanting to do an in depth analysis in the changes of particular OTUs then it doesn't help you separate out the real OTUs from the contamination.
There are also a bunch of other techniques out there which people are using. I am worried that researchers will end up selecting which method is used based upon the sequencing data they get back. Your extraction kit is full of Bradyrhizobium and not much else? Just take out all the OTUs. It's full of Pseudomonas? Just leave them them in then.
Anyone's opinion on what they think the best way to deal with is would be most appreciated.
Hi Laura, I think you sum up the options /dilemma excatly. And it must depend on what sort of question you are trying to answer. If you test for your contaminats from your blanc sequence controls, you will proberbly find that they are not really in your samples, since the PCRs run differently in samples from blancs and DNA containing ones so I wouldnt worry to much. Also you can design your studies in a way that the contaminants is any can be thought to have the same impact on your different experimental groups so as not to be a problem when analysing the results. In reality, you should always do this because of the risk of contamination is omniprecent.
Why can't you address this statistically? Compare both composition and levels of OTUs found in negative controls and test samples and subtract the former from the latter using error propagation?
Dear Laura,
Take a look at this paper: http://www.biomedcentral.com/1741-7007/12/87. Basically, there are several steps you can take to minimize the effect of contaminant OTUs on your analysis:
1. Maximize the starting sample biomass (by filtration or enrichment).
2. If PCR or extraction kit reagents are contaminated, replace them. If they are contaminated a second time, find another manufacturer.
3. Always sequence positive (mock) and negative controls from each batch of sample collection/storage medium, each extraction kit, and each PCR kit concurrently with the environmental samples of interest. Sequence the controls even if you don’t see a band on the gel.
4. If you have many samples, randomize or divided them so that treatment groups are evenly represented across all extraction kits and sequencing runs. Record which sample was processed with which kit so that contamination of a particular kit lot number can be traced to the final dataset.
5. After sequencing, look at taxa that are present in the negative controls, taxa that are statistically associated with a particular batch of reagents, and taxa that are unexpected biologically.
6. To qualify as "contaminants" in the samples, reads in the sample must be the exact sequence (or within the margin of sequencing error) that we find in the negative control (both reads being a “Pseudomonas” is not good enough). Furthermore, it should be present in a similar abundance relative to the other contaminants in the sample. So, if sequences A and B show up in a 2:1 ratio in the sample and in the control, it’s likely a contaminant; if it’s a 20:1 ratio in the sample and 1:2 in the control, it’s probably not a contaminant. I don't recommend “correcting” the relative abundance in the sample to remove the contaminant signal.
Hi Zohar,
Do you have a reference for your point#6? It makes sense what you say. It would be great if you can provide a paper where it has been done already.
Thanks a lot!
- Badges
- Science topic
- Similar topics
- Papers