Investigations into the Effect of Multiobjectivization in Protein Structure Prediction
Physics-based potential energy functions used in protein structure prediction are composed of several energy terms combined in a weighted sum. ‘Multiobjectivization’ — splitting up the energy function into its components and optimizing the components as a vector using multiobjective methods — may have beneficial effects for tackling these difficult problems. In this paper we investigate the hypotheses that multiobjectivization can (i) reduce the number of local optima in the landscapes, as seen by hillclimbers, and (ii) equalize the influence of different energy components that range over vastly different energy scales and hence usually swamp each other’s search gradients. The investigations use models of two real molecules, the alanine dipeptide and Metenkephalin under the Amber99 energy function, and consider hillclimbers with a range of mutation step sizes. Our findings support the hypotheses and also indicate that multiobjectivization is competitive with alternative methods of escaping local optima.
KeywordsLocal Optimum Multiobjective Optimization Protein Structure Prediction Gaussian Mutation Alanine Dipeptide
Unable to display preview. Download preview PDF.
- 4.Day, R.O., Zydallis, J.B., Lamont, G.B., Pachter, R.: Solving the protein structure prediction problem through a multiobjective genetic algorithm. Nanotechnology 2, 32–35 (2002)Google Scholar
- 7.Handl, J., Lovell, S., Knowles, J.: Multiobjectivization by decomposition of scalar cost functions. In: Rudolph, G., et al. (eds.) PPSN X 2008. LNCS, vol. 5199, pp. 31–40. Springer, Berlin (2008)Google Scholar
- 11.Jones, T.: Crossover, macromutation, and population-based search. In: Proceedings of the Sixth International Conference on Genetic Algorithms, pp. 73–80. Morgan Kaufmann, San Francisco (1995)Google Scholar
- 12.Knowles, J., Watson, R., Corne, D.: Reducing local optima in single-objective problems by multi-objectivization. In: Proceedings of the Congress on Evolutionary Multiobjective Optimization, pp. 269–283. Springer, Berlin (2001)Google Scholar