Train_test_validate function does not work for nci60 dataset

[jjacobson95]

The train_test_validate function currently fails when it encounters an NaN value such as in nci60.

Edit - for more info: This is still present even after dropping NaN values from each datatype using the following command:

cd_nci60.experiments.dropna(inplace = True)
cd_nci60.transcriptomics.dropna(inplace = True)
cd_nci60.mutations.dropna(inplace = True)
cd_nci60.copy_number.dropna(inplace = True)
cd_nci60.drug_descriptors.dropna(inplace = True)
cd_nci60.drugs.dropna(inplace = True)
cd_nci60.proteomics.dropna(inplace = True)
cd_nci60.samples.dropna(inplace = True)

[ymahlich]

This is a dataset problem. Those NaNs are generated because there is no fit_auc for certain entries in the nci60.experiments object. See attached screenshot.

This needs to be either fixed on the dataset / build side or, we need to exclude those experiments (which should probably also happen on the build side of things)

[sgosline]

This will be fixed when #340 is fixed.