ROC curve for metadata classifier

[LenaLapidot]

Hello and thank you for your great work!
I’m using the q2-sample-classifier and was wondering wether it is possible to get a ROC curve and a value of AUC?
The result I get include the confusion matrix and the feature importance heatmap.
Also, in the qzv of the heatmap, the legend is measured as log10 frequency - Can you please refer me to an explanation?

Thank you.
Best,
Lena

[Nicholas_Bokulich]

@LenaLapidot

I’m using the q2-sample-classifier and was wondering wether it is possible to get a ROC curve and a value of AUC?

I have added ROC/AUC to a development branch of QIIME 2... hopefully these will make it into the 2019.7 release — when that comes out next week, take a look at the changelog to see if this is one of the new features that made it in! If it's not in 2019.7, it will be released in 2019.10, or you can attempt to install the development version on your computer.

@LenaLapidot

Also, in the qzv of the heatmap, the legend is measured as log10 frequency - Can you please refer me to an explanation?

The colorbar indicates log10 frequency (how many times that read was detected). So if a feature was detected 100 times in a sample, log10(100) = 2, etc. I believe using the --p-no-normalize parameter will lead to raw frequencies being plotted.

Let me know if you have any questions!