NCBI or SILVA for curating a custom 16S rDNA reference database?

General Discussion
1

Hello all,

I'm seeking advice on creating a 16S rDNA database for an avian diet metabarcoding project. I chose the primers 16S1F-degenerate and 16S2R-degenerate (Deagle et al 2007) alongside COI ANML primers to generate 150bp paired end data. I am currently building custom reference databases for each of these primers; my study species is a generalist shorebird so my reasoning was that ANML would target insect prey while 16S would be able to capture other potential prey items like flatworms, crustaceans or fish for example. For the ANML set, I'm following this tutorial using get-ncbi-data. However, I am uncertain about whether I should use NCBI or SILVA when curating sequences for the 16S. My impression is that SILVA is more bacteria focused, and I read this thread where they find that SILVA collapses the taxonomy of some eukaryotes. However, in reading the SILVA manual, it seems more curated than NCBI. Overall, I'm curious what recommendations users who have worked with either datasets would have for creating a metazoan/animal reference dataset. Depending on your recommendations, I also have additional questions:

If I should go the get-ncbi-data route, do you have any pointers on how I can find a list of 16S alternative dictions to use as --p-query search terms? For COI, this tutorial had a nice list. Would something like this be sufficient?: --p-query "txid33208[ORGN] AND (16S OR 16S rDNA OR 16S ribosomal DNA)".

If I should got the get-silva-data route, I understand that the SILVA database is largely in rRNA format, so should I reverse transcribe the results of get-silva-data, then follow this tutorial as normal?

Thank you in advance for your time and help. I am a complete newbie to QIIME and metabarcoding so lengthy explanations are welcome!

Best,

Clare

1 Like
2

Hi @deClare125,

There are several ways you can go about this...

  1. For Eukaryotic specific reference sequences, you can try
    • get-eukaryome-data: use SSU to grab 16S rRNA gene data.
    • get-midori2-data: to grab mitochondrial reference data. Use srRNA for the 16S rRNA gene data, and CO1 for cytochrome oxidase subunit 1 data.
  2. Then, if you'd like, you can follow the extract-seq-seqments tutorial using these reference databases.

^^Note: the online documentation does not appear to reflect the current database options for these commands. So, be sure to check out the respective options using the --help flag.

I've not tested this, but given that you are are performing an eDNA diet survey, I am thinking that these two databases might be better tailored for assessing diet eDNA. Then you can use GTDB, SILVA, and Greengenes 2 to assess microbial diversity.

For the NCBI query, I'd try and limit it to non-microbial data using the taxonomy ID to restrict which taxonomic groups you like to keep or ignore. But I think the two databases I mentioned above have already done much of the hard work for you. So, give those a try and see how they work for you.

If you go with using get-silva-data I'd follow the initial steps to fetch the data from here, through reverse transcribing, then you can use either feature-classifier extract-reads alone, or work through the extract-seq-segments tutorial.

Please update us on how well these work! :slight_smile:

-Mike

3 Likes
extract-seq-segments: number of reads vs accuracy of reads.