| Protein crystals and crystallization are playing an ever increasing role in understanding protein function and protein-protein interactions, and also in the production of protein pharmaceuticals. The self-assembly of protein molecules into an ordered crystalline array is surely controlled by protein-protein and protein-solvent interactions. It is currently believed that molecular recognition, the geometric and chemical complementarity of protein molecules, is a predominant factor controlling these interactions, and hence, a protein's ability to crystallize, or crystallizability. However, a fundamental understanding of how surface interactions control a protein's crystallizability has not yet been realized. In this work we have validated a set of protein-protein interaction potentials for use in crystalline environments and have incorporated these potentials into a simulation of protein crystallizability. The simulation has been developed to account for atomic-level protein interactions to capture the interactions that control self-assembly and to account for the effect of mutagenesis.; An interaction energy test was developed to ensure that the potentials chosen were valid in crystalline environments. The design of the test was based on the hypothesis that a protein molecule in a crystalline environment should have a global minimum net protein-protein interaction energy when in the crystallographically expected orientation. The test systematically explored the net interaction energy of a protein molecule in all possible orientations at a fixed location within a crystalline lattice. The interaction potentials tested were a combination of knowledge-base united atom pair potentials and a soft steric repulsion, and the global minimum orientation energy for two different proteins was within 5° of the crystallographicalIy expected orientation. This fact provided confidence that the potentials were suitable for our protein crystallizability simulation. Also, the interaction test correctly indicated the existence of a protein crystal polymorph for one of the systems.; The simulation of protein crystallizability was a lattice-based Grand Canonical Monte Carlo simulation, where the protein molecules were modeled as rigid united atoms. The simulation investigated the phase behavior of arrays of protein molecules as a function of chemical potential. The dimensions of the lattice and the coordinates of the united atoms were taken directly from the Protein Data Bank file of the system being simulated, giving a physically realistic representation of the protein monomer.; The simulation was validated by simulating five protein crystal systems with known crystallization behavior. The correct crystalline behavior was observed in four of the systems. The simulation indicated that the proteins studied exhibit crystallizability in specific ranges of chemical potentials, analogous to a “crystallization slot” (George and Wilson, 1994), which we have defined as a crystallizability window. The simulation was able to correctly rank the anecdotal crystallizability of lysozyme, insulin, and maltose-binding protein. We also studied the crystallization behavior of four maltose-binding protein mutants. The simulation correctly indicated that the different mutant forms of maltose-binding protein preferred their natural crystal lattice to the lattices of the other maltose-binding protein mutants. |
