Yu Xia,

i worry you've posed a question which is either to difficult or to delicate to answer. I've followed this question now for all seven days and yet no valuable answer to this interesting and profound question were given - sigh!

Albeit metagenome assemblies are giving me headaches and I'm generally critical about them i can't add anything qualified to this discussion as it is not my key research field.

As i would also like to read more about this by other researchers i feel to bump it up.

- Sebastian

I m working with metagenomic bacteria from wastewater. I never tried to make any draft genome since its nearly impossible, I have million of different lines making their assembly nearly impossible. Depending on your work area and metagenomcis sequence origin, you "might" try it. Sorry if my answer if not really helpful.

Loubna

Hi Youssar,

I would suggest a good example to retrieve draft genome from metagenome (though there sample is enriched sample so may not as complex as real wastewater) in the below paper. Hope it may raise your interest :)

Albertsen, Mads, Philip Hugenholtz, Adam Skarshewski, Kåre L. Nielsen, Gene W. Tyson和Per H. Nielsen. “Genome Sequences of Rare, Uncultured Bacteria Obtained by Differential Coverage Binning of Multiple Metagenomes”. Nature Biotechnology advance online publication (2013). doi:10.1038/nbt.2579.

Dear Yu Xia,

a recent and impressive example IMHO for the reconstruction of population-level genomes from metagenomic data is the work by Nielsen et al. (2014): Identification and assembly of genomes and genetic elements in complex metagenomic samples without using reference genomes. There, they report the recovery of "238 high-quality microbial genomes" from fecal metagenomes. The human gut is nowadays often considered as "well-characterized" in terms of bacterial organisms living in there. "Well-characterized" meaning here that extensive efforts have gone into this while it is at the same time of "amenable" complexity.

If you consider for example metagenomics of soil that is a completely different story. A very recent paper from Howe et al. (http://www.ncbi.nlm.nih.gov/pubmed/24632729/) states for example the following: "more than half of the assembled contigs were not similar to any sequence in the MG-RAST m5nr databases, suggesting that soil holds considerable unexplored taxonomic and functional novelty". And we are talking here about de novo assembled data where "80% of the sequencing data could not be assembled because of low coverage". Arguably, tackling such highly diverse environments will hardly be feasible using culture-dependen approaches, hence the need for efficient and accurate computational tools.

Regarding metagenomics vs. single-cell sequencing, I would consider these approaches complementary. Microbial communities are known to exert a certain amount of "gene trafficking", meaning that you have transfer of genetic elements from one cell to another. Accordingly, no two cells will be alike! While small variations (e.g., SNVs) may not be a big problem, some cells may have acquired genetic elements while others not. An example of this would be commensal microorganisms that acquire pathogenic factors. You see that there is quite some genetic heterogeneity there and hence a single-cell approach will be able to give you a snapshot of a few cells, but those will only be a few. Going to the extreme and assuming no experimental challenges (there are a lot though!) doing single-cell sequencing of _all_ the cells in a microbial community would probably allow maximal resolution. But I would say that this is something for the not-so-near future. Hence, metagenomics can be used to build draft genomes which might then be refined for example using targeted single-cell sequencing based on specific markers designed from the draft genomes (I am thinking along the lines of FISH).

You can also view this as an answer to your question on how accurate a population-level reconstructed genome can be (well, actually rather how one could assess how good the reconstruction actually is) and how, theoretically, you could "convince" a reviewer.

Coming back to your point on the value of a draft genome. You are absolutely right that it could be used to discover necessary growth conditions to guide isolation, e.g., http://www.ncbi.nlm.nih.gov/pubmed/23417797 or http://jb.asm.org/content/191/18/5877.full .

Oh, any finally, something out of personal interest, I would _not_ consider population-level genome recovery using genomic signatures like tetranucleotide frequency _no_ challenge anymore. They still offer their challenges! Approaches, like the one from Nielsen2014, may have been very successfull also in the automated recovery, but they require several samples and powerful comouting resources among other things. This comes of course at a price. It is not only the sequencing costs but also in the first place having enough samples, which is not always easy. Genomic signature-based approaches can work on single samples. Speed, sensitivity, accuracy, and ease-of-use are important factors on which we recently improved considerably (http://www.nature.com/srep/2014/140331/srep04516/full/srep04516.html ). And I would not say that we have reached the end yet there.

Best,

Cedric