Hello Gerard,

Your question doesn't really have a simple answer, as it depends on a couple of variables. The most important being what kind of phylogenetic resolution you are looking for.

If a genus abundance characterization is enough, i think you can get a good estimation using around 10.000 reads per sample. (see "A core gut microbiome in obese and lean twins", Turnbaugh et al.).

However, if you're looking for characterization of the low abundant species in the tail of the abundance distribution, saturating that might not be that easy. Check out "Human gut microbiome viewed across age and geography" (Yatsunenko et al.), where they do Illumina hiSeq on the 16S to get about 1.8 million 100 bp reads per sample. You can do a rarefaction analysis on that to find a cutoff that is suitable for your purpose.

Hope this is helpful.

This is a not such an easy question to address generally. For the chip size it depends on how many samples you have. For a single sample even a 314 should be more than sufficient. If you want to compare different samples, then dependent on the amount you might need to go with the greater chips. The main problem lies here in the down-stream processing. Dependent on how stringent you define and filter out low quality reads you need to adjust the per-sample output of the sequencer. My personal suggestion would be (and please don't pin me down on this) to make sure to have at least ~50k reads per sample after all filtering etc. Given the knowledge that for some sequencing projects the part of removed reads can even make up more than 50% of the raw data I would not sequence more than 10 samples on a 316 chip. Having the option to apply very stringent filter criteria while still remaining in good sampling coverage is a reliable starting position for such projects and such complex microbiomes.

For the targeted region, I worry there is not that much scope left. 200bp is not that long and they also need to span the multiplex ID and sequencing adapters. As primer selection is very critical and also you should consider the habitat under study I can not give any concrete suggestions. Generally combining different adjacent regions results in a higher phylogenetic resolution, but even the shorter used combinations are still too long (e.g., V1-V2, V3-V4, V6-V7).

all are having marginal ideas only including me 40Mbp of 200bp data is enough

Hi Gerard!

I was on a similar situation some months ago. After giving a few rounds, we decided to use the 515F/805R primers for V4 region (http://www.pnas.org/content/early/2010/06/02/1000080107).

They have very nice coverage for Bacteria and Archaea (take a look to suppl. data), so you can look both kingdoms.

These primers are thought to be used with Illumina system, so you should modify adapter/linker sequences to fit IonTorrent system.

Although the objective is amplify a 250 bp paired-end fragment, the last part is not so informative, so with first 150bp from the 515F is enough to get accurate phylogeny. Actually, using paired-end-reads (250bp) or forward-reads (150bp) give the same results.

Regards.

Hello Gerard,

In our lab (mostly human gut microbiome samples), as a rule, we sequence the V1V2 region, going from 3' to 5' end, from the BSR357 primer. We have tested v1v3 and v3v5 regions (from from 3' to 5') and got similar results as well. Sequencing from the V9 region into the 16S gene seems to yield the most divergent results (perhaps not that unexpectedly).

This paper (from Rob Knight's group) was published relativelly early on the pyrosequencing days (when 454 was still FLX and had only ~250bp) and is really helpful in selecting what 16S region to work with:

Short pyrosequencing reads suffice for accurate microbial community analysis.

Liu et al. NAR. 35(19):e120.

http://www.ncbi.nlm.nih.gov/pubmed/17881377

As for depth of sequencing, we usually shoot for ~5000 sequences per sample for general surveys of intestinal microbiome samples (mind you, the number is not set on stone, and sometimes even as little as 1000 sequences or less can be enough - depending on the question at hand).

One of the reasons we shoot for 5000 is that even though we try to normalize our sample pools, sometimes, the the number of sequences recovered can vary a bit, and this way chances are we'll be able to use all samples even if we have to rarefy them down.

The depth of sequencing needed is usually a consequence of the effect size you are interested in. As Paul mentioned, if you must see that supper rare bug, you'll likely need to get many more sequences per sample.

This paper shows (and others of course) show nicely that a very high depth of sequencing is often unnecessary:

Direct sequencing of the human microbiome readily reveals community differences.

Kuczynski et al. Genome Biology. 11(5):210.

http://www.ncbi.nlm.nih.gov/pubmed/20441597

Hope it helps,

Chris

The use of Ion torrent for 16S rRNA gene based metagenomic analysis is a difficult task at this point in time as the technology is new and most of the softwares available for data analysis are not tailor made for this tech. We use a 316/318 chip for human fecal microbiota estimates and there is a lot of sequence loss (quality filtering) when we analyze the data. For a scientifically sound data representation the multiplexing of 16 -20 samples for 316chip should give you enough data for estimating microbial diversity at reasonable depth. The level of multiplexing is upto the user and the goal of the study. QIIME and Mothur have to be integrated in DATA analysis. The 200 bp chemistry would give you mean read length of around 150-170bp. so filtering the reads is necessary as there will be alignment issues, and you will have to try various window range for your sequences. The homo-polymer issue has to be kept in mind during analysis. finally the selection of variable region is challenging. use of V3 gives good diversity representation but equimolar pooling of this is very tediuos (multiple peaks in BA [3 to 5]) and concentration estimates will be approximate while pooling. It is better to use V6 if there is difficulty in equimolar pooling.

The use of Ion torrent for 16S based metagenomic analysis is challenging but once you are successful in standardization; it is fun to work with.

All the best for your study.

I am analysing a quite simple microbial sample associated with decaying plant material. Using 454 sequencing of two different amplicons performed on 16S regions (V3-V4) and (V5-V6) the results looks like very similar. With 5000-10000 reads per sample and rarefaction analysis the microbial population seems to be nearly saturated. The results obtained are at the genus level and agree quite well with a previous sampling performed on cultivable bacteria isolated from the same sample. Did you consider to perform random sequencing (not only on 16S) using ion torrent?

@ Stefano: metagenome (WGS) sequencing on Ion torrent has a major limitation with respect to coverage and depth. A low complexity ecological niche such as decaying plant material or say rock communities can be done on 318 chip which in most cases would give you 700 to 900 Mb of data. There is a need to modify your DNA extraction protocol to minimize contamination from other sources and get around 90 to 95 % of microbial DNA so that the low depth of sequencing can be counter balanced to some extent (one sample on one 318 chip). We have not done WGS metagenome as we deal with a highly complex community (Human gut) and prefer illumina.

Well you can always give a try, after all new avenues have to be explored in research. There is one group that has tried this approach ref: http://www.sciencedirect.com/science/article/pii/S0167701212002345