1rst: Even using the entire SSU rRDNA sequence, it is not always possible to resolve better than the genus.

2 nd: when using only a smalll variable domain, it can be worse.

3rd: I am not familiar qith QUIIME, since I have my own pipeline. When I want to have more resolution, I extract the reads for this genus, recover the refseq sequences for this genus, align and do a tree. This is long and difficult, but this is presently the only way; and often you need help form a specialist of this genus

Best regards

Richard

Hi,

it is quite difficult to distinguish 2 species using 16S since they may differ in only several nucleotides of their whole 16S and you have just part of the gene (V3/V4 regions with your primers). So, you actually lost much information during library preparation, not during QIIME analysis. Using 16S V3/V4 should not be overestimated, it work pretty well, but genus level is maximum to provide reasonable information. If you want to get more information on species level, you should probably change completely your sequencing strategy to get better "resolution". In your current data, you don't have the information about species, I guess.

QIIME is the best software for metagenomics study. I think there is some problem with sequencing. You can even try with MGRAST, which is an online freeware.

Approaches using percent-similarity binning often have resolution problems. For higher-resolution, you might want to check out oligotyping (http://merenlab.org/projects/oligotyping/) or MED (http://merenlab.org/projects/med/).

Hi Karel，

  Thanks for you information.

    What do you mean by 'change completely sequencing strategy to get better "resolution'? Thanks!

Hi,

  Thanks for all your message on this issue.

   For primer set (341-806), it will give about 430 bp sequence, and I know this will loss many many information on species level. But I want to get much more information on species, that is tricky for me.

  For QIIME part with assign_taxonomy.py, there  are five assign reference, which one should I use for assign taxonomy to get better results?

1) uclust consensus taxonomy assigner (default)

2) SortMeRNA

3) BLAST

4) RDP Classifier

5) RTAX

Thanks!

Hi Xin,

it's hard to say. Depends on results you expect. If you want primarily qualitative study, just to identify species in your sample, I would prefer shotgun sequencing. You would be able to assemble reads into longer contigs that should contain enough information for species determination. However, I don't have personal experience with such an approach. If you want also quantitative study, to find out species abundance, I don't have any idea. NGSs don't have satisfactory read length.

HI Xin.

I would say that the method for assigning taxonomy is not as important as selection of the database against which you are comparing. In our studies we use BLAST, but we do say against many databases and use an LCA method of resolving taxonomy. Whilst it can give species resolution, you should not expect to species resolution for many bacteria as the databases often cant resolve that far.

This dpends on many factors such as clustering level, databases coverage, BLSAT stringency, primer selection (and bias).

dont be surprised if the best you can get is genus level identification. If you need species level then you need to really consider your primers compared with the genus' that you want to get species level id from.