go back to Research Group home page
go back to Complex Systems home page

 
 

 Protein folding by means of fast Markov Chain Monte Carlo algorithms

(Berg, Grassberger, Hsu, Mehra)


The determination of the natural state of proteins is maybe the most demanding and most important problem in computational biology. Traditionally, physicist are more in favour of Monte Carlo method (MC), in view of their greater flexibility. The main difficulty of the protein folding problem arise from the fact that the potential landscape with many false minima , similar to the energy landscape in a glass. In recent years, an increasingly large number of efficient of MC methods have been proposed for dealing with this kind of problems. 

Apart from using PERM to analyze proteins, we also use modern versions of Markov Chain MC methods. At the moment we use mostly parallel tempering, but we plan to use also other methods. The public domain package SMMP are used for reading in the force field parameters and the coordinates of the particles, calculating the energy, etc. There are three force fields, ECEPP/2, ECEPP/3, and FLEX implemented in SMMP,  which bond lengths and bond angles are held fixed.

Lattice models (HP models) in 2D and  3D are studied by using various sampling strategies of PERM.
Similar strategies are used also for protein-folding in a continuous space, and their performance is  compared.