Input for MelonnPan

Hi friends!!! @franzosa @himel.mallick @fbeghini @f.asnicar
I want to use melonnpan for my data and I have gone through the tutorial. I have two questions:
Firstly, my HUMAnN3 genefamilies output is in the RPKs unit. How should I convert it so that it can be used as an input file for MelonnPan? The tutorial says it is in terms of proportion. Then why the sum is not equals to 1.00 for the test dataset?
Second, I am working with control samples and obese samples. I don’t have any metabolome data that can be used for training purpose. In that case, can I use the pre-trained model?

Thanks
DC7

Hi @DEEPCHANDA7 - the test dataset of the MelonnPan tutorial has been preprocessed to contain the most informative features - in other words, following TSS normalization, we have removed features that have low variance based on prevalence and abundance. Hence the sum is not 1.

Regarding your specific application, you can use the pre-trained model with the obvious caveat that the default model was trained in an IBD cohort which is presumably different from your study population.

All the best,
Himel