An interesting paper published in the J. Chem. Inf. Mod. investigates the role of water molecules in CDK2 and CDK5 active sites. The authors use MD simulations to simulate water and inhibitors (in this case the well-known indirubin analogs) in the active site of the CDKs. The interesting observation they make is that in some cases, the inhibitor "recruits" a water molecule to form a bridged interaction with an atom in the protein, thus achieving the "correct" binding mode. In other cases, blockage of the active site by a loop prevents this water grabbing trick.
The study also highlights how it can make a big difference between using monomeric CDK and CDK bound to a cyclin when simulating nib ("nib": inhibitor...apparently new yippie slang) binding. In this case, the difference is in the water recruitment.
Again, one is reminded of the many subtle factors, including water interactions, that dictate differences between nib binding to different but similar kinases.