Less number of total Species than Genus in 16S data

Question

[arshad1292]

Hello,

I analysed the 16S data using the DADA2 pipeline with standard parameters. After ASV classification, the data show total number of species less than total number of genus. Is this normal or not? I was assuming species should always be more than Genus but confused with the results I obtained. Any thoughts please?

Many thanks in advance!

Answer 1

[Philipp Bayer]

DADA2's classifiers might not be able to classify to the species-level, it will keep the taxonomic level 'higher' as necessary.

Example with some of my fish ASVs:

"Eukaryota","Chordata","Actinopteri","Labriformes","Labridae",NA,NA  

Only family-level classification, no genus/species

"Eukaryota","Chordata","Actinopteri","Kurtiformes","Apogonidae","Apogonichthyoides","Apogonichthyoides timorensis"  

Classified to the species-level

So yeah, you might have more genera classified than species if your reference database contains many very similar species. It's likely!

[arshad1292]

What's the solution to that? Should I try qiime2 or kraken2 instead? that may give deeper classification to species level?

[Philipp Bayer]

It really depends on what your marker gene is, what your biological question is etc.!

In my case, I often simply can't go to the species level. For example, in Lutjanidae (snappers) the 16S gene is sometimes completely identical across all species so no classifier in the world will be able to tell where that particular sequence came from, it will assign the genus or family level.

You can save some species-level assignments by manual curation of your reference database. If you have four hits of species A in family A, and one hit of species B in family B, the taxonomic classifier will go to the next highest taxonomic level. Curation means removing the likely misannotated species B. But this is not an easy process, it involves much manual work.

You can try Kraken2, in my experience it will assign many false positives unless you run it with --confidence 0.05, but I doubt that it will magically find species-level labels where before it found nothing. Might be better to find a reference database that has been manually curated or is specific to your region of interest (like meta-fish-lib is specific to UK fish).