A numerical study of hybrid optimization methods for the molecular conformation problems
An important area of research in computational biochemistry is the design of molecules for specific applications. The design of these molecules depends on the accurate determination of their three-dimensional structure or conformation. Under the assumption that molecules will settle into a configuration for which their energy is at a minimum, this design problem can be formulated as a global optimization problem. The solution of the molecular conformation problem can then be obtained, at least in principle, through any number of optimization algorithms. Unfortunately, it can easily be shown that there exist a large number of local minima for most molecules which makes this an extremely difficult problem for any standard optimization method. In this study, we present results for various optimization algorithms applied to a molecular conformation problem. We include results for genetic algorithms, simulated annealing, direct search methods, and several gradient methods. The major result of this study is that none of these standard methods can be used in isolation to efficiently generate minimum energy configurations. We propose instead several hybrid methods that combine properties of several local optimization algorithms. These hybrid methods have yielded better results on representative test problems than single methods.
- Research Organization:
- Sandia National Lab. (SNL-NM), Albuquerque, NM (United States)
- Sponsoring Organization:
- USDOE, Washington, DC (United States)
- DOE Contract Number:
- AC04-76DR00789
- OSTI ID:
- 10175387
- Report Number(s):
- SAND-93-8233; ON: DE93017898
- Resource Relation:
- Other Information: PBD: May 1993
- Country of Publication:
- United States
- Language:
- English
Similar Records
Quantum Monte Carlo Endstation for Petascale Computing
Real-time fast inverse dose optimization for image guided adaptive radiation therapy-Enhancements to fast inverse dose optimization (FIDO)