Currently the separation of predictors into bpp, mfe, mea, and pk often does not make much intuitive sense. Eg there are predictors in mfe that can predict PKs and there is repeat functionality between mea and pk. I would suggest reconsidering these separation and creating more intuitive groupings.