An improved parallel differential evolution approach for protein structure prediction using both 2D and 3D off-lattice models

Protein structure prediction (PSP) is a well-known problem in bioinformatics. Identifying protein native conformations makes it possible to predict its function within the organism. Knowing this also helps the development of new drugs and the comprehension of some biological mechanisms. During years some techniques have been developed for this purpose but, due to their high cost, it is necessary to use simplified models of protein structures. However, even the simplest models, with low biological plausibility, are excessively complex from the computational point of view. This paper reports the application of Differential Evolution (DE) to solve the PSP problem using a Toy Model (also known as the AB Model) in both 2D and 3D to represent the protein structure. This work presents two different versions of the DE algorithm (basic and adaptive) using a parallel architecture (master-slave) based on Message Passing Interface in a cluster. Some special operators for DE were developed: explosion and mirror mutation. All tests executed in this work used four benchmark sequences, ranging from 13 to 55 amino acids. The results for both parallel DE algorithms using both 2D and 3D models were compared with other works in the literature. The DE algorithm achieved excellent results. Overall results encourage further research towards the use of knowledge-based operators to improve further the performance of DE.