It is not general, you have to decide it yourself based on similarity/identity analysis of your gene/genome.

To add to what @AbhijeetSingh said,

Sometimes we find high % similarity genes/proteins, only to later find out they are non-functioning homologs.

So the way we approach it is by building a list of sequences from microbes that have been shown to have a functioning homolog of the particular gene/protein we are looking for. Then we can use this list as a reference database to blast our sequence data against.

Hope that helps!

I agree with Abhijeet Singh and Sanjin Mehić. Mostly scientists use nucleotide identity (DNA or RNA sequences) to define OTUs from metagenomic/metatranscriptomic data sets, because already available databases, e.g., NCBI GenBank, Ribosome database, etc. are mainly based on nucleotide sequence identity, so, if you want to analyze your data based on protein similarity, there are also limitations regarding already available data.

Sanjin Mehić Salma Mukhtar Abhijeet Singh

thank you for your answers.

You convinced me that I need to use nucleotide and not protein sequences to run the clustering.

Nevertheless, these are RNA viruses, not microbes strictly speaking.

Sanjin Mehić I am not sure if I understand your strategy, could you explain it in more detail?

@Guillermo I was just trying to say that you should double check the OTUs at some point before performing extensive analyses on them.

Sometimes when i manually blast an OTU, it comes back as identical to a version that has been experimentally shown to be inactive in a biological context. So even though the identity truely is similar, its not significant in the environment because its not active.

I would do protein identity and not nucleotide identity for all protein-coding genes. In the end, the protein is the conserved code, not the nucleotide sequence. Both for DNA and RNA viruses. I would say, what is your opinion?