Post Fri Sep 14, 2018 7:24 am

Python phenograph classification

Dear CyTOF forum,

I am a student from the Netherlands who just started an internship regarding CyTOF data analysis. Among other methods I am exploring phenograph and have successfully run phenograph clustering on an AML dataset which gives communities but these in itself are fairly meaningless. I've plotted the unclustered data on two TSNE dimensions with its original labels:

Image

Now I wish to use phenograph to train on part of my data and predict labels for the other part so I can compare phenograph's classification with the ground truth labels, and here I ran into problems. I cannot figure out how the input for the phenograph.classify is supposed to be formatted. I'm using the Python implementation as found here https://github.com/jacoblevine/PhenoGraph

Does anyone here have experience with this module, and could you tell me in which format to supply the data to the phenograph.classify() method?