Because the predictive models are trained with
five-fold cross-validation, there are five different models
for each predictive problem, each with different coefficient values. Therefore, the feature importance of a gene
on a particular predictive task is the average normalized rank of that gene’s regression coefficient across the
five cross-validation folds. 



Reaseons that why the important features occurs on edges?

edge features are those feature has low varience

these feature points have distinct intensity in the patients among other featurepoints