Predicting the Initial Steps of Salt-Stable Cowpea Chlorotic Mottle Virus Capsid Assembly with Atomistic Force Fields.

Computational prediction of native protein-protein interfaces still remains a challenging task. In virus capsids, each protein unit is in contact with copies of itself through several interfaces. The relative strengths of the different contacts affect the dynamics of the assembly, especially if the process is hierarchical. We investigate the dimerization of the salt-stable cowpea chlorotic mottle virus (CCMV) capsid protein using a combination of different computational tools. The best predictions of dimer configurations provided by blind docking with ZDOCK are rescored using geometry optimization with the Amber and Rosetta force fields. We also evaluate the relative stabilities of the three main interfaces present in the icosahedral capsid using locally restricted docking with Rosetta. Both the rescoring and locally restricted docking results report a particularly stable protein-protein interface, which is the most likely intermediate during the first stage of the hierarchical capsid assembly. The blind docking results rescored with both Amber and Rosetta yield docking funnels, i.e., three or more near-native structures among the top five predictions. The results support experimental observations on in vitro assembly of CCMV capsids. The cross-validation of the results suggests that energy-landscape-based methods with biomolecular force fields have the potential to improve existing docking procedures.