In this paper, we present an extension of our phenotype algorithm for global cluster geometry optimization from atomic clusters to the more difficult molecular clusters. A successful application to the benchmark case of TIP4P water clusters shows that the new method is at least as good as one of the best methods in the literature so far. We point out why the TIP4P water cluster system starts to become very complicated beyond twenty water molecules and follow the traces of a beginning structural transition from clusters with all water molecules at the surface to clathrate-like cages with a truly interior water molecule.