Free energy calculations in globular proteins: Methods to reduce errors

Abstract

The calculation of free energies by computer simulation represents one of the most promising areas in molecular modeling. While the computational methods developed so far give reliable results for liquids or solutions, they are not satisfactory for globular proteins. The reproducibility of the data is poor due to several sources of error. The most important are due to the magnitude of the molecule's phase space, to the long relaxation time of the system, and to the singularity occurring when creating or annihilating atoms. In a previous study Simonson and Brunger reported the free energy differences calculated for three ¨ successive mutations in the ribonuclease-S system and revealed several sources of error. These errors were reanalyzed and the performance of several methods studied in order to reduce them. Different approaches of mutating the Hamiltonian are compared using the method proposed by Resat and Mezei and a modification of the method proposed by Cross. Procedures are also proposed to reduce the effects of the long relaxation time of the molecule, to bias the simulation toward the experimental structure, and to reduce large free energy derivative fluctuations. All these methods give reliable results when the mutation is carried out in a peptide in solution. When the mutation is carried out in a globular protein, the sources of errors are reduced but not eliminated. Although the investigated procedures and methods increase the reliability of free energy calculation, further improvements will be required. Q 1998 John Wiley & Sons, Inc. J Comput Chem 19: 1229)1240, 1998