Premium
Conformational memories and a simulated annealing program that learns: Application to LTB 4
Author(s) -
Guarnieri Frank,
Wilson Stephen R.
Publication year - 1995
Publication title -
journal of computational chemistry
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.907
H-Index - 188
eISSN - 1096-987X
pISSN - 0192-8651
DOI - 10.1002/jcc.540160512
Subject(s) - simulated annealing , monte carlo method , computer science , statistical physics , minification , sampling (signal processing) , phase space , algorithm , energy minimization , importance sampling , mathematical optimization , theoretical computer science , mathematics , physics , statistics , computational chemistry , chemistry , filter (signal processing) , computer vision , thermodynamics , programming language
The use of computer simulations in all areas of chemistry is growing rapidly because of the powerful insights that they have provided into many interesting phenomena. As investigators continuously examine more sophisticated problems, they need increasingly more powerful tools. Hence, much effort has gone into the development of algorithms which might extend the scope and power of standard dynamic and Monte Carlo techniques. In the Monte Carlo regime, the most common area subject to improvement is the choice of a trial move. In the ordinary case, trial moves are generated uniformly at random. In the extended and hopefully improved case, trial moves are generated randomly but not uniformly. In this article we present a new and totally general method of biased sampling which is applicable to any flexible molecule. In our method, multiple simulated annealing runs are performed to reveal populated and unpopulated regions of the multidimensional conformation space. The second phase of the simulation is done at a fixed temperature with sampling only from populated regions found in the first phase. Because the simulated annealing runs quickly reveal unpopulated regions of the conformation space, the volume of conformation space that needs to be sampled in the second phase of the algorithm is reduced by many orders of magnitude. Additionally, because no energy minimization is used, these populations represent a canonical ensemble which may be used to estimate conformational free energies. © 1995 by John Wiley & Sons, Inc.