Hi all,

Apologies if this is very long winded, first time post on this forum.
I am using protein data to predict disease status, using WEKA to model my predictions with various algorithms. I get very good sensitivity and specificity but I would like to do now is construct some cut off points for my proteins to see where the optimal level is.
Since I have several proteins that combine in the model to give the best prediction. How will I go about applying some form of cut off to see if my Sens' or Spec' increases? I'm a bit stuck on how to format my data set in order to do the classification.
Would I recode my protein values into high and low, if so do I do that on my whole population or just my training dataset?

Any advice and help will be greatly appreciated.
Thank you