Extract Files Sample-Classifier


I am using the sample-classifier plug-in to see if a treatment impacts the microbiome, which we have bench evidence to support. Our training vs. testing group had strong correlations.

Is there a way to parse out the populations that are predicted to correlate with a specific treatment? I know that it requires a lot of validation, but we are wanting to do that on the bench and we want to know a bit more about the populations identified by this plug-in.

Hi @Kmf,

I think you want the FeatureData[Importance] artifact from classify-samples or the more granular actions. Note that this won’t give you a population, but it will tell you which features are more or less informative to the classifier.

(@Nicholas_Bokulich, please correct me if I’m wrong here :smiley: )

1 Like

As @ebolyen suggested, look for the feature importance results. If you are using the latest version of QIIME 2, this will be one of the output artifacts. If you are using an older version, the important features appear at the bottom of the visualization along with importance scores.

1 Like

This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.