I plan to do some 16S targeted NGS (most likely paired end reads with MiSeq) to study microbial communities in wetland soil samples. I am looking for some guidance to help me decide how many reads I need per sample. My research question does require that I see the "rare" organisms as much as possible, but I need to balance that with cost. I have a very large number of samples and must multiplex. I would appreciate any guidance and/or suggestions for new articles I should consider. Also - I assume I need fewer reads to characterize Archeae than Bacteria but I do not know of any articles that specifically discuss this.
This is a million dollar question that haunted me two years ago (and it still does). Give a thought on why do you need to do NGS while you could just do clones or may be DGGE/TTGE. Soil systems are usually highly diverse and so the rarefaction curves hardly reach the plateau. Hence, when it comes to rare biosphere, you will need superbly high number of reads to get any reasonable picture. Since such "rare" reads are not of your interest multiplexing with a few thousand reads might do for you. As for my work, after normalisation I ended up with c2000 reads per sample (120 samples) which was enough to show the difference and identify the dominant ones.
If you do find this info useful and want to know more, feel free to ping me on my inbox. I can pass on draft of my paper where we did such multiplexing and got reasonably good resolution.
We find 10,000 sequences per samples i generally sufficient, but as stated above, depends on your diversity, so it depends what kind of soil you are looking at as well. do you have any initial idea of what the diversity is from other published work?
We had a range of between 4134 and 9611 amplicons per sample. This was sufficient to use for statistical analysis. It is part of the Earthmicrobiome project. You can look up all the data on their website and data from other projects too.
Sanjay is right. Microbial ecology, as a field, is trying to get a better understanding of this very question. It really depends more on the hypothesis you are trying to test. The "rare" microbes are hard to survey as valid sequences because of technical error in NGS/library prep giving you false taxa. On the other hand, their importance may be of value to your project. Akin to wolves in a forest, rare but important, the rare microbes could have a large impact to the ecology of the soil. A lot of what I have seen in the literature has had sufficient reads/sample around 2K-10K, much like what others have posted here. If you have the funds to do a pilot sequencing run, or can throw in a few samples to another sequencing run you would be able to estimate the number of sequences that would be sufficient for your samples and hypothesis more confidently when you do your full analysis.
Feel free to keep this discussion going on this thread as you come up with more questions.
Hi Rima :) - check out Caporaso JG et al. (2011) Colloquium Paper: Global patterns of 16S rRNA diversity at a depth of millions of sequences per sample. Proc. Natl. Acad. Sci. U.S.A. 108:4516–4522. Good luck with the analyses!
I would second Robert's response recommending a pilot run with a few samples. We do a lot of 16S amplicon based diversity assays in soils, and most are done in the 3-10k range. However, given your interest in "rare" organisms, you really need to do a pilot run and look at rarefaction curves to tell if you are getting adequate coverage. The insight you glean from these rarefaction curves will be a good investment - especially given that you have a large number of samples.
And yes - Archaea typically do have lower diversity.
Hi! In microbial ecology numbers can be misleading. First of all capturing "rare" species is challenge considering the currently available technologies. The reason for this is at the studying Microbial diversity using amplicon sequencing come with its own bias. As per my experience, selection of primers is a major concern. You will have to look for primer set that has least bias in amplification. The "rare" seems to be the ones that have not been successfully identified and databases lack these sequences, they are filled with sequences from the dominant groups. So first you check out the primers that give you least bias in amplification. Optimizing pcr conditions is also required so that you avoid any biases inherent pcr bias eg. number of cylces, annealing temperature, etc. Ref: The long-term stability of the human gut microbiota. Faith JJ, Guruge JL et al 2013. Science. 2013 Jul 5;341(6141):1237439.
Lee CK, Herbold CW, Polson SW, Wommack KE, Williamson SJ, et al. (2012) Groundtruthing Next-Gen Sequencing for Microbial Ecology–Biases and Errors in Community Structure Estimates from PCR Amplicon Pyrosequencing. PLoS ONE 7(9): e44224. doi:10.1371/journal.pone.0044224
As mentioned by Kristen, the paper by Caporaso JG et al. (2011) does give some good idea about the importance of using mock communities as these help as 'controls' for the sequencing run and help in sample data analysis. I agree with Robert that sequencing errors hamper the confident taxonomic assignment of reads, so mock communities help in these aspects. One more thing, it is always better to check it for yourself the best parameters. In which case you run one sample that you expect to be diverse with different primer sets and pcr conditions along with a mock community, get yourself accustomed to the analysis and then you will be able to take a decision. This approach will take some time initially but will help you design your study better. Feel free to contact for further queries. All the best with your work!!!
In my optinion it is not possible to suggest a number of ampilcons for detection of rare organisms in soil samples without any future information about the sample (what kind of wetland, which physico-chemical conditions, …). Like Stephen Cox I would recommend preliminary runs with few samples to get an idea about the community. And second important point is you find only what you looking for. That means you need a well aligned primer set, as mention Sudarshan Shetty before.
Hi Rima :) - I'm happy to read of you. I would reccommend to consider Lemos et al.(20119 Rethinking microbial diversity analysis in the high throughput sequencing era.J.Microbiol methodsw 86,42-51.All the best for you
- I'm happy to read of your question. I would reccommend to consider Lemos et al.(20119 Rethinking microbial diversity analysis in the high throughput sequencing era.J.Microbiol methodsw 86,42-51.All the best for you
There have been a lot of recommendations for the rarefaction test, which is a good thing for you to do, but if you get some pilot data. However, seeing as you are working with soil microbes, there is a chance that the rarefaction will never plateau. Soil samples are notoriously diverse. That being said, it would at least give you a better estimate of the total sequences you that would be best for each sample when you do a larger experiment.
The alternative would be to look around the literature for an article that has some similarities with your sample set and use their data as a litmus for the approximate number of sequences you will need for your own work.
Hi Rima
Yes, an interesting question indeed. Other contributors here have given some good advice and interesting references. I'll add Caporaso et al PNAS 2010 into the mix. They used Illumina on a variety of samples, generating over 3 million reads per sample. In depth analysis of the resulting reads showed that 2000 were sufficient to preserve the sample relationships observed with the full data set. Of course the rare biosphere is another question, you'll probably need substantially more than this. I would think however that running 96 samples per MiSeq run would get what you need, giving you a ballpark figure of 200,000 reads per sample, depending on conditions.
As for the Archaea/Bacteria question, why not use a universal 16S primer set (as in Caporaso et al above, for example). This will negate the need to worry about differences in abundances during sample prep.
Good luck.
It has been shown that about 2000 reads per sample is enough to distinguish effects of environmental factors as 100k reads.
http://www.pnas.org/content/early/2010/06/02/1000080107
You don't have a lot of control with Illumina and 454 as the number of reads obtained per sample depends on the sequencing efficiency of each amplicon pool as well as amount added to the plate. Another problem is that the bias in sequencing coverage/reads per sample generally increases as you add more samples to your lane. We have sucessfully sequenced a pool of ~100-120 samples in one Illumina lane, but some of my colleagues tries to limit the number to 50 or so samples per lane. There are fairly good primer sets out there that can provide coverage for both domains.
I'm presently working on a similar topic and working on this question too! I'm not looking for rare species exactly but looking at the functional diversity of microbes. 6 years after your post, I'm still wondering on sample size that will be representative of the practical scenario! thanks to your question I have got some really good answers.
- Badges
- Science method