We investigate the use of a Mayer cluster expansion to quantify changes to the energy landscape of an ab initio protein structure prediction Hamiltonian as side-chain excluded volume radii are increased. The expansion is generalized so that the effect of other changes in the Hamiltonian, which would alone be best treated with a cumulant expansion, can be evaluated simultaneously. The effect of changing the excluded volume radii on various thermodynamic quantities, in particular the free energy and glass transition temperature, is calculated with the expansion and compared to results from simulation. The expansion is found to be sufficiently accurate at the second-order level to be useful for calculating values of the excluded volume radii that refine the energy landscape for structure prediction. This is explicitly checked by a free energy profile analysis and examining the quality of predictions before and after the optimization. A useful method to test whether equilibrium has been attained during molecular dynamics sampling is also described.
Read full abstract