Premium
Optimization methods for conformational sampling using a Boltzmann‐weighted mean field approach
Author(s) -
Huber Thomas,
Torda Andrew E.,
van Gunsteren Wilfred F.
Publication year - 1996
Publication title -
biopolymers
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.556
H-Index - 125
eISSN - 1097-0282
pISSN - 0006-3525
DOI - 10.1002/(sici)1097-0282(199607)39:1<103::aid-bip11>3.0.co;2-h
Subject(s) - boltzmann constant , generalization , sampling (signal processing) , chemistry , boltzmann machine , molecular dynamics , boltzmann distribution , statistical physics , basis (linear algebra) , field (mathematics) , computational chemistry , physics , mathematics , thermodynamics , computer science , artificial neural network , mathematical analysis , artificial intelligence , geometry , detector , pure mathematics , optics
An optimization protocol is proposed that combines a mean field simulation approach with Boltzmann‐weighted sampling. This is done by including Boltzmann probabilities of multiple conformations in the optimization procedure. The method is demonstrated on a simple model system and on the side‐chain conformations of phenylalanines in a small hexapeptide. For comparison, calculations were performed using classical stochastic dynamics simulations [M. Saunders, K. N. Houk, Y. Wu, C. Still, M. Lipton, G. Chang, and W. C. Guida (1990), Journal of the American Chemical Society, Vol. 12, pp. 1419], iterative optimization of probabilities on a fixed set of basis conformations [P. Koehl and M. Delaure (1994), Journal of Molecular Biology, Vol. 239, pp. 249–275], and simulations with locally enhanced sampling [A. Roitberg and R. Elber (1991), Journal of Chemical Physics, Vol. 95, pp. 9277–9287]. Although approximations are used in our method, the results may be more physically meaningful than those of the other procedures discussed. Furthermore, the approximate Boltzmann distribution allows generalization of the results. © 1996 John Wiley & Sons, Inc.