Hi Stefano,

your question is not so clear to me. Metagenomic approach has the advantage that is more "representative" of the present community. Unfortunately, it is extremely difficult to reach the sequencing depth you need for it. On the other hand, targeted specific sequencing (16S & 18S PCR) will give you easier the necessary depth, but you are introducing the PCR amplification bias. A third approach would be to do untargeted unspecific metatranscriptome sequencing (meaning sequencing the total RNA extracted) that will give you high sequencing data on 23/16S and 28/18S RNA. There you will have the most active "players" of your community, but you will still have problems regarding the sequencing depth. If you are interested in the community composition in genus level (anyway, for environmental samples does not really making sense to go to species level) the best would be bidirectional amplicon sequencing (preferable in 454 Pyrosequencing platform). Especially for environmental samples, dont care so much for the "deep biosphere" (if this is not your target) but for the species really contributing to the processes...If by chance a species is there in extremely low copies, it will probably not do much...Of course if you are working in deep biosphere, I will happily like to read your future papers ;-)

@Henrik Nielsen. Computational resuorces and bioinformatic skills are not a problem. We are thinking about random sequencing to have a better representation of the microbial community and, if possible, to have a first glimpse on the genomes of the more abundant microorganisms. We are concerned that most of the random sequences aligned on the eucariotic genomes and only a small portion on the prokaryotis genomes.

The requirement for computational resources and skills are a different issue. Another factor you need to consider is the sequencing costs. The question is how many samples you wish to analyze. If the number of samples are many and your goal is just to characterize the taxonomic composition and diversity of the different samples (or the same environmental sample at different temporal points), a 16S rDNA sequencing should be your choice.

Using the MID-based multiplexing options in current sequencers, you can sequence multiple 16S rDNA samples in a single run of sequencing. You can get a comprehensive picture of the microbial community composition with minimal investments in sequencing. In contrast, for random DNA sequencing, to obtain the complete picture of the taxonomic as well as functional composition, you will have to go for one sample per run which may lead to a massive increase in the sequencing costs.

On the other hand, if the number of samples to analyze is few, you can as well go for random DNA sequencing (ideally one sample per run using say the 454 ones), this approach will not only enable obtaining a good enough picture of the taxonomic composition but also facilitate a deeper investigation into the functional aspects.

I have been able to assemble large sections of genomes (even completing some) from shotgun metagenomic sequence - however, this was using DNA from an animal that had a few major symbionts (along with many other minor ones). If your DNA is from a very complex environment you may not be so successful, or you may need very deep coverage. We've been basically doing one HiSeq 2000 lane for each sample. Having said that, you may be able to get a less biased picture that way. There are programs like EMIRGE that can extract full length 16S sequences (which may be better for classification versus amplified sections) from metagenomic paired-end Illumina data. In my hands EMIRGE has given me sequences that were later confirmed by Sanger to be accurate, but I'm not sure how accurate it is for low minor members of the microbiome.

Hi Stefano,

your question is not so clear to me. Metagenomic approach has the advantage that is more "representative" of the present community. Unfortunately, it is extremely difficult to reach the sequencing depth you need for it. On the other hand, targeted specific sequencing (16S & 18S PCR) will give you easier the necessary depth, but you are introducing the PCR amplification bias. A third approach would be to do untargeted unspecific metatranscriptome sequencing (meaning sequencing the total RNA extracted) that will give you high sequencing data on 23/16S and 28/18S RNA. There you will have the most active "players" of your community, but you will still have problems regarding the sequencing depth. If you are interested in the community composition in genus level (anyway, for environmental samples does not really making sense to go to species level) the best would be bidirectional amplicon sequencing (preferable in 454 Pyrosequencing platform). Especially for environmental samples, dont care so much for the "deep biosphere" (if this is not your target) but for the species really contributing to the processes...If by chance a species is there in extremely low copies, it will probably not do much...Of course if you are working in deep biosphere, I will happily like to read your future papers ;-)

@Michael Granitsiotis: Do you have more informations on the sequencing of untargeted unspecific metatranscriptome sequencing? I think the sequencing of 23/16S and 28/18S RNA to identify the more active species could be very useful. Do you know if someone is using this approach?

I subscribe completely the answer of Henrik Nielsen and add PCR introduces strong biases and makes any quantitative approximation highly unreliable. Shotgun sequencing is better even if you are only after taxonomy that in the case of prokaryotes is quite inert anyway.

It might be too late but I would like to suggest the following publication:

Shakya, Migun, Christopher Quince, James H Campbell, Zamin K Yang, Christopher W Schadt, and Mircea Podar. "Comparative Metagenomic and rRNA Microbial Diversity Characterization Using Archaeal and Bacterial Synthetic Communities." Environ Microbiol 15.6 (June, 2013): 1882-99. Print.

Basically, they found, using a synthetic community, that Metagenomic outperform amplicon sequencing which is introducing significant taxon-specific biases influenced by the primer set used.

@Simon Labrie

Thank you for your suggestion, very interesting paper!

i recommend you go for MLSA analysis like rpoB gene which is always present in single copy number in bacterial genome and more reliable than 16S rRNA for bacterial identification at species level. be happy

Hi Stefano

Well the 16S sequence analysis has some artifacts such as;

1. some of the bacteria have more than 1 kind of 16S different sequences.

2. Horizontal gene transfer can also occur.

These problems and some other will mostly lead to false data, however at the moment we don't have any solution to overcome these. So we just go for it, hope in future we may have some alternatives, but now we have to rely on this Oldy goldy method.

Well in your case I will suggest to go for 16S rather than random sequencing, as your aim is just to determine the diversity at genus or specie level rather than functional one.

Best of luck.

Hi..

I did 16sRNA sequencing for the bacterial community recently. I got a good read length for all samples and from the some of the samples I sent for different sequencing company showed same genus but different species. Later I found that it occurs mostly in Bacillus group. But for my study thats fine .Now I am doing for metagenomics appraoch for another objective in this study. My suggestion is you may go for 16sRNa for culture isolates and if you need more isolates that non-cultureable, metagenomics approach may be the opt for you.Recently I found your question, so sorry for the dealy..:-)

I would like to add to the discussion, that it also depends on the environment you want to sample and analyze. Very diverse environments like soils and carbon rich sediments have a very high diversity. Shotgun sequencing in those cases requires deep sequencing to get enough coverage to be able to do sequence assembly, which would make the metagenomics step more interesting.

If you are only interested in the diversity, you could chose to filter out all reads matching rRNA /rpoB sequences, and trash the remaining.

The advantage of amplicon sequencing over shot-gun sequencing is that you can process a lot more samples in one run, which helps in setting up a proper experiment. Although shotgun sequencing outperforms A.sequencing, it's not so easy and cheap to process a lot of high diversity samples.

rRNA was needed in the '90s. It provides very poor resolution and most of the time helps very little in understanding the community. Metagenomics of total DNA (or maybe single cell genomics when widely available) are the only way to make head way. Computer power or no computer power available. TIME TO SAY GOOD BYE TO rRNA, WE ARE IN THE GENOMICS ERA.

In addition to previous work mentioned, take a look at this paper. Also supported Metagenomic sequencing is more accurate than 16S rRNA amplicon.

Logares, Ramiro, Shinichi Sunagawa, Guillem Salazar, Francisco M. Cornejo-Castillo, Isabel Ferrera, Hugo Sarmento, Pascal Hingamp, et al. “Metagenomic 16S rDNA Illumina Tags Are a Powerful Alternative to Amplicon Sequencing to Explore Diversity and Structure of Microbial Communities.” Environmental Microbiology (2013): n/a–n/a. doi:10.1111/1462-2920.12250.

@Yu Xia

Thank you very much! I missed completely this very intersting paper in environmental microbiology. This is exactly what I was looking for.

Many important aspects have already been discussed, and I would mostly agree to the answers stating that shotgun metagenomics provides a richer, less biased view on your community. However, even though many people believe that 16S amplicon sequencing will soon be an obsolete technology, under some circumstances 16S sequencing still has an edge over shotgun metagenomics in my opinion:

• In our settings, shotgun metagenomics (for a diverse community as the human gut) is at least a factor of 10 more expensive than 16S amplicon sequencing (using Illumina platforms). If sample acquisition is cheap, you could also rephrase your question as a trade-off between sample size and "richness" of the sequencing read out. For statistical reasons you might be better off with 16S in some settings (e.g. case control studies or comparisons between environments or over time) as you have more power to detect subtle differences between microbial communities. Given the immense variability of many microbial communities, power issues are often overlooked.
• When interested in sampling host-associated communities you often have to deal with "contamination" with host DNA, for example when you try to scrape off microbes from the mucosa of colon biopsies. While this can pose a problem for shotgun sequencing (as you mostly get human reads rather than bacterial DNA you're after), the 16S primers will only amplify bacterial DNA and you won't spend most of your budget on "useless" reads. You can find mention of this e.g. in the supplement of Kostic et al. or a discussion in the context of metatranscriptomics in Westermann et al. (see links below).
• When talking about quantification bias, people often mention PCR amplification bias, whereas it is frequently ignored that different sample handling, DNA extraction methods, or computational analysis methods can strongly bias your results too (I haven't seen a study that tried to compare their impact). You could have a look at Fig 3b in Sunagawa et al. (attached below) to get an idea how much of an impact different DNA extraction protocols can have.

Article Genomic analysis identifies association of Fusobacterium wit...

Article Dual RNA-Seq of pathogen and host

Article Metagenomic species profiling using universal phylogenetic m...

Hi Stefano!

This is really dpends on your aims. IF you want to describe OTUs content, distribution withour any functional description the 16S RNA-targeted approach is very useful. For functional metagenomics it can be useful full sequencing but is also very expensive and I do not think useful for diagnostci purposes.

Hi  Stefano

This is depends on you. If you interested to know the whole community of microbes in any environmental samples then you go with the 454 sequencing.

We have recently published a paper regarding differences in the anaerobic digestion microbiome determined by the sequencing strategy Article Taxonomy of anaerobic digestion microbiome reveals biases as...

We evidenced differences in abundance of specific taxa due to shotgun sequencing, 16S amplicon and direct sequencing of the 16S rRNA transcript.

An intersting discussion on the same topic was started by Holly Bik in twitter, if you are interested see https://twitter.com/hollybik/status/1013081253782831104