Using an aggregation tree to arrange energy function terms for protein structure prediction

In many real-world problems, as the protein structure prediction (PSP), a number of conflicting objectives have to be simultaneously optimized. In this paper, the Aggregation Tree (AT) method is applied to arrange the energy function terms used by a protein three-dimensional structure prediction program (called GAPF) that is based on a multiobjective genetic algorithm. The results achieved using the aggregation provided by the AT method are compared with those obtained in a previous study, in which the arrangement of the objectives was performed using only the knowledge of experts, without the aid of any aggregation tool. We observe that the AT method is a promising aggregation strategy for problems involving the selection and the arrangement of many objectives, having suggested an interesting arrangement of the GAPF energy potentials, with results equivalent to those obtained using expert knowledge.