Clustering of metabarcoding reads from many environmental samples

Question

[lvogel]

Hi,

I've been using CD-HIT EST in the past to cluster sequences of COI from environmental samples. For my current project, however, there are over 75 samples, and my supervisor wants to know if all of the sequences from all of the samples can be pooled, and then clustered, and then in the end we can know which sequences came from which sample. I tried searching, and I found that CD-HIT OTU does something like this with pooled sequences from multiple samples, but it's apparently only for 16s rRNA, and I have only eukaryotic organisms (COI). Could anyone recommend a clustering program for me?

Thank you for your time,

Laur

[lvogel]

Hello gb,

Actually, I did end up moving on to USEARCH for most of the steps of my metabarcoding pipeline. It worked well and required only minimal scripting knowledge. Recently, however, I've been tasked with even larger datasets, which exceeded the 32-bit memory of the free version of USEARCH. So I'm currently learning how to use VSEARCH. You're correct, though, and should get accepted answer.

[gb]

Yes, I am in the same situation. With VSEARCH you can also make an otu table but with the lack of documentation it is difficult to find the right commands. Look at this page: https://github.com/torognes/vsearch/wiki/VSEARCH-pipeline

Answer 1

[gb]

Hello,

This answer comes a bit late but you can exactly do this with usearch.

https://www.drive5.com/usearch/manual/pipe_otus.html

https://www.drive5.com/usearch/manual/cmd_cluster_otus.html

If you prefer CD-HIT above uparse you need to do some extra steps for otu clustering like length trimming, find uniq read abundance and sort on the read abundance. Using CD-hit alone is not the proper way to do otu-clustering.