(151b) A New Multiscale Algorithm and Its Application to Coarse-Grained Peptide Models for Self-Assembly

Scott Carmichael, Rafael Prato, and M. Scott Shell

Department of Chemical Engineering, University of California, Santa Barbara

Email: carmichael@engr.ucsb.edu shell@engr.ucsb.edu

Peptide aggregation plays a role in a number of neurodegenerative diseases, such as Alzheimer's, Huntington’s, and Parkinson’s.  Here we aim to develop an accurate molecular-scale picture of  the aggregation process using a multi-scale computational approach.  Recently, Shell [1] developed a coarse-graining methodology that is based on a thermodynamic quantity called the relative entropy; a measure of how different two molecular ensembles behave.  By minimizing the relative entropy between a coarse-grained (CG) system and an all-atom (AA) system, an optimized coarse-grain model can be obtained.
Here we develop a corresponding numerical strategy for optimizing CG models of arbitrary peptide sequences, and apply it to a model polyalanine molecule (Ala)15.  Specifically, we optimize the CG force-field parameters that control the behavior of one, two, and three bead per amino acid models of (Ala)15.  Simulations of the optimized models peptide are performed, and histograms of various structural correlations (e.g., bond length, bond angle, etc.) are extracted and compared to an AA system.  We obtain excellent agreement in these distribution functions for all of the optimized force-fields, indicating the CG models are able to accurately reproduce much of the behavior of the original AA system.  Moreover, the models yield single-molecule folding curves and free energy surfaces whose accuracy improves with the level of model detail.  
We subsequently use these CG models to simulate the self-assembly of large-scale peptide systems that are too computationally demanding to study using conventional AA simulations.  Simulations containing many CG (Ala)15 peptides form a β-sheet structure with excellent agreement with the  experimentally-observed [2], structure typical of alanine rich peptides.We use these large-scale simulations to provide a detailed picture of the self-assembly mechanism and the effects of chain length and concentration.

[1] Shell, M. S. J. Chem. Phys. 2008, 129, 144108-7.
[2] Forood, et al.,  Biochem. and Biophys. Research Comm. 1995, 211, 7-13.