The knowledge about the conformation of a protein molecule allows the inference and study of its biological function. Because protein function is determined by its shape and the physio-chemical properties of its exposed surface, it is extremely important to predict accurate protein models. One of the hardest problems in Structural Bioinformatics is associated with the prediction of the three-dimensional structure of a protein only from its amino acid sequence (primary structure). Coils and turns are both elements of secondary structure in proteins where the polypeptide chain reverses its overall direction; These structures are considered the most difficult secondary structure to be predicted. In this paper, we propose a loop Structure Pattern Library (SPL) which was created using experimental information extracted from Protein Data Bank aiming to constrain the conformational search space of proteins. The Self-Adapting Differential Evolution (SADE) meta-heuristic was implemented for the tertiary protein structure prediction problem using the Structure Pattern Library as knowledge. The SADE algorithm was tested with (SADE-SPL) and without the Structure Pattern Library. Archived results show that the lowest Root Mean Square Deviation values were obtained when the Structure Pattern Library was employed. Average GDT_TS were higher in all SADE-SPL cases. Thereby, our results allow us to state that SPL application knowledge in SADE meta-heuristic is capable of predicting three-dimensional protein structures closer to experimental structures than SADE application without SPL.
