Coordinate generation of negative examples and cross validation folds between network prop and PPI prediction

[tmmurali]

For network propagation, we plan to augment host interactors (of viral proteins) from the Krogan dataset with predicted host interactors. Since we will perform cross validation for both tasks, we have to coordinate generation of negative examples and cross validation folds between them so that no information "leaks" from one task to the other.

[tmmurali]

@jlaw9 Can we close this issue? Which part of the code handles it?