i am trying to come to grips with the QIIME2 taxonomic assignment options, and 'feature-classifier ’ doesn’t have a much explanation for how it works. It has as methods, BLAST and SEARCH, then also the trained Naive Bayesian classifier. Is it using all of these? Is it using them in sequence? What are the benefits of using this plugin?
Is there some simple explanation for how this works? If I wanted to use it for AMF primers against MaarjAM would this be straight forward to ‘train’?
Thanks for posting!
The short answer is that the
classify-sklearn method is essentially a python version of the popular RDP classifier, which is another Naive Bayes Classifier. Their performance is very similar when using default parameters, but see the publication link below for more details.
The short answer is yes, training a new classifier is fairly straightforward, theoretically on any marker gene (so long as your database is formatted correctly). This process occasionally raises errors when running on a new database and formatting is usually the issue; if you get an error please search this forum and this error may already be described by previous users. I am not familiar with MaarjAM but I’m guessing that you are using ITS as a marker? We use the UNITE database for ITS, so you can check out that resource as a “template” for formatting the MaarjAM database for your needs.
These are all separate methods that you must call separately. Each is described in the publication link below.
classify-sklearn is the naive Bayes classifier similar to RDP, as described above. BLAST and VSEARCH use the alignment algorithms of the same names to find top hits in a reference database, and then find consensus taxonomies among these N top hits. More details in the link below.
Short answer: their performance is equivalent or superior to other commonly used sequence classification methods. See this preprint to learn more about these methods and their performance relative to other common amplicon sequence classifiers. Results for 16S and ITS (which I assume you are using) are shown.
I hope that answers your questions! Happy QIIMEing.
This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.