Ye Xiang, Wang Jun, Luo Ray
Department of Molecular Biology and Biochemistry, University of California, Irvine, California 92697-3900.
Department of Molecular Biology and Biochemistry, University of California, Irvine, California 92697-3900 ; Department of Biomedical Engineering, University of California, Irvine, California 92697-3900.
J Chem Theory Comput. 2010 Apr 13;6(4):1157-1169. doi: 10.1021/ct900318u.
A revised density function is developed to define the molecular surface for the numerical Poisson-Boltzmann methods to achieve a better convergence and higher numerical stability. The new density function does not use any predefined functional form but is numerically optimized to reproduce the reaction field energies computed with the solvent excluded surface definition. An exhaustive search in the parameter space is utilized in the optimization using a wide-range training molecules including proteins, nucleic acids, and peptides in both folded and unfolded conformations. A cubic-spline function is introduced to guarantee good numerical behavior of the new density function. Our test results show that the average relative energy errors computed with the revised density function are uniformly lower than 1% for both training and test molecules with different sizes and conformations. Our transferability analysis shows that the performance of the new method is mostly size and conformation independent. A detailed analysis further shows that the numerical forces computed with the revised density function converge better with respect to the grid spacing and are numerically more stable in tested peptides.
开发了一种修正的密度函数来定义分子表面，用于数值泊松-玻尔兹曼方法，以实现更好的收敛性和更高的数值稳定性。新的密度函数不使用任何预定义的函数形式，而是通过数值优化来重现使用溶剂排除表面定义计算的反应场能量。在优化过程中，利用参数空间中的穷举搜索，使用包括折叠和未折叠构象的蛋白质、核酸和肽在内的多种训练分子。引入三次样条函数以确保新密度函数具有良好的数值行为。我们的测试结果表明，对于不同大小和构象的训练分子和测试分子，使用修正密度函数计算的平均相对能量误差均统一低于1%。我们的可转移性分析表明，新方法的性能大多与大小和构象无关。详细分析进一步表明，使用修正密度函数计算的数值力在网格间距方面收敛性更好，并且在测试肽中数值上更稳定。
