I want to understand how well a particular tissue site (tissue_site1) reflects another tissue site in the same patient. I want to employ sample-classifier as a sort of sourcetracker, training it on tissue_site1 then runing the classifier on location gradients of tissue_site2 to see how it affects the proportion predictions. 1. Is this a valid approach? 2. Is there a way I can feed the training and testing feature-tables separately into the qiime sample-classifier classify-samples
command so I can do something like this?
I am not entirely sure what you mean here — training on one tissue type and then testing on another tissue type will probably just lead to poor predictions but perhaps I misunderstand your approach.
Yes, the classify-samples
pipeline outputs a trained classifier that can be re-used, e.g., with a new dataset.
Better yet, don't use the pipeline, use the individual methods to train a classifier on one dataset and then test on another... check out the q2-sample-classifier tutorial at qiime2.org for a flowchart of the individual steps involved to get a sense of how to piece together a custom pipeline.
Good luck!
1 Like
This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.