thanks for coming back to me with these very helpful comments. I basically spent the last days following up on your suggestions. The vsearch classifier seems to be a good option and I am running it currently on 5 threads (also using SILVA’s 7 level taxonomy instead of all). Though, it has not concluded yet after 2 days… Now another post made me think that something else is wrong with my data: (Merged sequence 100%OTU picking). Our Illumina data was merged with Casper and demultiplexed, quality-checked outside of q2. So, in the post I learned that paired-end pre-joined sequences cannot be used in the qiime 2 workflow prior to 2017.11 is that right? Might that explain my large amounts of sequences and the inability of the classifier to conclude in reasonable time? Also, I am having trouble with alignment using MAFFT as well which is quickly running out of memory (just in case this might be related).
What would be the best way forward in order to account for the pre-merged status of my sequence data using the new 2017.11 release?
Thanks so much!
steffen