Hi,
I'm doing metabarcoding analyses with eDNA samples and after clustering my sequences (97 % similarity) and did a blast (97 %) with a reference data base (UNITE), I notice that some OTUs have the same species identification. Logically they should have been grouped together during the clustering step.
Thank for your help
First, the BLAST algorithm only compares the similarity between your OTUs and whatever is available in your database. So, you need to check the similarity cut-off, percentage match etc.
Second, since your database is somehow incomplete, BLAST tries to assign your OTUs to the nearest related taxa in the database. It doesn't mean all taxonomic assignments are correct.
I suggest you try other robust taxonomic assignment methods
https://besjournals.onlinelibrary.wiley.com/doi/abs/10.1111/2041-210X.12399
http://www.drive5.com/usearch/manual/taxonomy.html
https://github.com/rdpstaff/classifier
Hi Genevieve, I agree with the answer above.
Your sequences can be >3% different to each other and therefore clustered in separate OTUs. However, when these OTUs are blasted, a taxonomic assignment is made based on sequences available on the dataset you are using (e.g. GreenGenes; NCBI ). Sometimes the taxonomic assignment is "forced" to the closest match (which, paradoxically, could be only 80-85% similarity). This is why your OTUs, while >3% divergent between each other, are assigned the exact same "taxonomical ranking".
How to deal with this (in my opinion):
Always remember to check the matching % of your OTU with the assigned taxonomy. If necessary, manually blast it yourself and check how similar is the most similar sequence. If your OTU is only 80-90% similar to the most similar sequence available on the dataset, this should be what you have to report and to keep in mind.
For example, an OTU 85% similar to a (e.g.) Sodalis glossinidius sequence is NOT a Sodalis glossinidius sequence. It probably belong to the same genus or family, but it would be a big mistake to report it as that species..
Hope this help!
Hello Genevieve,
this is perfectly normal and a good situation. The original sequences (or reads) may easily be less than 97% different, but globally similar to a same biological species AFTER clustering. Then you only need to merge them after biological species assignation.
Then again, if your pipeline would merge several biological species into same OTU, THEN you would have a bigger (but still rather small in big picture) problem, since it would not be easy to tear them apart. Then you would need to do the clustering again with a higher (98%) threshold.
I agree with the above answers that you may try other methods, like SINTAX etc. I have easy scripts to pull sequences from databases (like BOLD) and automatically turn them into SINTAX database, for example. It would need some work to make it compatible with fungi, but for them I would suggest this:
Article Protax-fungi: A web-based tool for probabilistic taxonomic p...
- Badges
- Science method
- Similar topics
- Biological Science
- Ecology