How to build the greengenes equivalent db for 12S rRNA

[Cornel]

Hi,

As the subject says, I need to create the equivalent of greengenes ref database, but for 12S.

I’ve looked at qiime2 documentation and it’s not very explicit (https://docs.qiime2.org/2017.8/tutorials/feature-classifier/ ).

The 1st step was to get all needed mitochondrial genomes ( https://www.ncbi.nlm.nih.gov/genomes/GenomesGroup.cgi?opt=organelle&taxid=2759 ) and their taxon ids.

Next, I should build the OTUs and taxonomy. Here I think I should extract the 12S region from each genome, but I’m not quite sure how to approach this.

Once I have the OTUs I can use scikit-learn to build the classifier.

I’m quite new at this. Any hints would be appreciated.

Thank you

[Nicholas_Bokulich]

Hi @Cornel,
Sounds like quite an ambitious project.

@Cornel

I’ve looked at qiime2 documentation and it’s not very explicit

q2-feature-classifier assumes that you already have a fasta file of sequences ready to use for database construction. It does not document how to re-generate greengenes because it does not assume (nor require) that level of effort. You could simply provide it with a fasta of full-length 12S gene sequences and let it do the rest. @BenKaehler anything to add?

@Cornel

The 1st step was to get all needed mitochondrial genomes

You could also just pull 12S gene sequences from genbank, instead of full mt genomes.

@Cornel

Here I think I should extract the 12S region from each genome

That's described in this section of the tutorial. You just need primer sequences for either end of the domain of interest.

@Cornel

Once I have the OTUs I can use scikit-learn to build the classifier

OTUs will speed things up but are not necessary. You could do this as quickly and easily as just downloading all full-length 12S sequences from Genbank and following the tutorial instructions.

If you really want to create the 12S equivalent of greengenes, you should contact the greengenes authors about that — they did a lot more that the steps you describe, particularly for handling taxonomies. When collapsing into OTUs you need to decide how you handle conflicts in taxonomy labels between different members of a cluster. So just using full-length 12S reference sequences should be adequate and easier, actually, than the steps you propose.

Hope that helps. Let us know if you have any further questions.