Sharing the Software / Models Used in Predicting 16S rRNA Copy Number

[Roland Wilhelm]

Hello there!

I'm building a pipeline to predict the 16S rRNA copy number from draft genome assemblies. The copy number posted to IMG/ER for my test set of genomes is in agreement with many of my predictions, and seems to be more accurate than my own (since many of my predictions are 0 or 1 for organisms we'd expect to have high copy number). So, I tried recapitulating your results using RNAmmer as described in Markowitz et al. (2014). However, these predictions are also discrepant (typically underestimate) with what is posted on IMG/ER in many cases. Would it be possible to provide more detailed information about what parameters / software is being used in the IMG/ER pipeline. I used the vanilla RNAmer HMMs in what could be in accordance with this text from Markowitz: "Ribosomal RNA genes (5S, 16S and 23S) are predicted using hmmsearch against the custom models generated for each type of rRNA in bacteria and archaea (,)."

Thanks in advance for any assistance you can provide,

Roli

[Marcel Huntemann]

Hi Roland!

Yes, the current pipeline runs hmmsearch, but it runs it against an in-house created set of models that got created by a few of the analysts here.

The complete pipeline proceedings are described here:

https://www.ncbi.nlm.nih.gov/pubmed/26512311

and for metagenomes here:

https://www.ncbi.nlm.nih.gov/pubmed/26918089

We are in the process of switching to a new pipeline though and that pipeline runs actually cmsearch against the Rfam models.

Thanks,

Marcel