I've followed the posts on the forum that Silva-128 was released for q2-fragment-insertion because so much work had already been done on it. Would it be valid to use Silva-132 for q2-feature-classifier and Silva-128 for tree construction? I suspect that there might be a few miss-matches with the different versions (recently added bacteria may not have a direct home on the tree, or the tree distance might not correlate as expected with taxonomy in some bacteria), but would this still be preferable to using Silva 132 for taxonomic classification and de novo tree construction? Was the benchmarking with Silva-fragment-insertion compared to fast tree ever published or reported?
Thank you to the developers for all of their hard work!
Hi @bkayser,
I’m not involved with either SILVA or fragment-insertion just my 2 cents here. You can certainly create taxonomy and a trees separately using different databases as these are independently called in QIIME 2. The issues you mentioned may be valid but since you filter your feature-table to remove features that were not inserted onto the tree I don’t think you’ll run into any downstream issues. I would say insertion tree based on a slightly outdated version of SILVA is still better than using de novo approach, that being said as far as I am concerned only the greengenes database has been benchmarked and evaluated thoroughly. @Stefan might correct me if I’m wrong on this.