Searching for the global energy minimum in molecular clusters is a pretty hard task. There are several global optimization strategies that I know of: multiple starting points, simulated annealing, basin-hopping. What else is out there at the market?
The verachem vm2 ( http://www.verachem.com/products/vm2/ ) calculates protein-ligand binding affinities, although I have not used this software and I don't know how it works.
That is an interesting application, Vicente. However, I am more interested in the mathematical optimization methods, that search for global configurational/conformational minima.
I've found basin-hopping in GMIN to be pretty good for molecular clusters (a small amount of my code is in GMIN, so I'm not completely impartial here). You could also look at the OGOLEM evolutionary algorithm (www.ogolem.org).
There are also evolutionary algorithms and other methods for global optimization of molecular clusters. There is a recent code (EA_MOL) from our group that has been already employed in several cluster systems. It is available to download from: http://apps.uc.pt/mypage/faculty/qtmarque/en/software
Well-known methods are
Improved Billard Simulation (Graham & Lubachevsky 1998), Modified Billard Simulation (Szabo & Specht 2007), in general see Locatelli & Schoen "Global optimization for atomic cluster distance geometry problems", in "Distance Geometry", pp. 197–212. Springer, 2013.
Furthermore, there are lattice search and Monte Carlo methods (Northby 1987), NLP reformulations (Gockenbach et al. 1997), Simulated Annealing (Coleman et al. 1994), two step iterations (Takeuchi 2006), stochastic global optimization (Locatelli & Schoen), etc.
There is also the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) which is implemented in many languages:
https://www.lri.fr/~hansen/cmaes_inmatlab.html
And there is quantum annealing from DWave, but I haven't looked at the details of how they emulate spin and tunneling. Real quantum tunneling has an extremely low probability. Quantum annealing search algorithm sounds like an interesting choice for some problems in that the jump from the current local minimum may provide a feasible next start.
There is also a list of global search algorithms at:
http://www.mat.univie.ac.at/~neum/glopt/software_g.html
Different algorithms have different means of providing functions to describe
your objective.
- Badges
- Science topic
- Similar topics
- Chemistry
- Theoretical Chemistry
- Computational Chemistry
- Molecular Dynamics