Evaluating the Peptide Structure Prediction Capabilities of a Purely Ab-Initio Method

Peptides, such as hormones and antibiotics, play many biological functions. Peptides` tertiary structures are of paramount importance for understanding their function, as well as the interactions with other molecules. DEEPSAM^[1] (Diffusion Equation Evolutionary Programming Simulated Annealing Method) is a novel purely ab-initio global optimization algorithm aimed to predict the structure of peptides and proteins, from amino acid sequence, without any preliminary assumption. This method is an Evolutionary Programming algorithm whose mutation operators are built by hybridizing the advantages of three well-known optimization methods - Diffusion Equation Method (DEM), Molecular Dynamics Simulated Annealing (MDSA), and the BFGS quasi-Newton local minimization method.

In principle, DEEPSAM requires much less computing resources and is much faster than other structure prediction methods, such as Replica Exchange Molecular Dynamics (REMD), and therefore seems to be more suitable for industrial use.

The goal of this study has been to further evaluate the structure prediction accuracy of the DEEPSAM algorithm by running it against a large number of known NMR structures of linear peptides (10-20 residues), in an aqueous environment modeled by the GBSA implicit solvent model.

[1] M. Goldstein, E. Fredj, and R. B. Gerber: "A New Hybrid Algorithm for Finding the Lowest Minima of Potential Surfaces: Approach and Applications to Peptides", J. of Computational Chemistry, vol. 32, pp 1785–1800, (2011).