Riemann, Nico;Zacharias, Martin - Reversible scaling of dihedral angle barriers during molecular dynamics to improve structure prediction of cyclic peptides and protein loops

ECCB 2002 Poster sorted by: Author | Number

Next | Previous poster (in order of the view you have selected)

Title: Reversible scaling of dihedral angle barriers during molecular dynamics to improve structure prediction of cyclic peptides and protein loops	P134
Riemann, Nico; Zacharias, Martin nriemann@imb-jena.de Jena Centre for Bioinformatics (JCB), Theoretical Biophysics, Institute of Molecular Biotechnology, Beutenbergstr. 11, D-07745 Jena.

Peptide cyclisation is frequently used to restrict the conformational freedom of a peptide for example to enhance its capacity for selective binding to a target receptor molecule. Structure prediction of cyclic peptides is important to evaluate possible conformations prior to synthesis and if it can adopt a desired conformation as a preferred low energy state. The situation is similar in case of protein loop structure prediction where one is interested in identifying favourable structures for peptide segments of a protein. Due the conformational constrains imposed by cyclisation or due to the boundaries on the ends of protein loops low energy conformations of cyclic peptides and protein loop regions can be separated by large energy barriers. We employ molecular dynamics (MD) simulations combined with a potential scaling method to specifically lower the barriers for dihedral angle changes of a peptide segment. The approach consists of several consecutive simulations starting with very low dihedral energy barriers and reduced non-bonded interactions between atoms separated by three atoms followed by gradually scaling the potential until the original barriers are reached. Application to several cyclic peptide test cases indicates that lower energy states can be reached by the potential scaling method compared to standard MD simulations. In addition, the relaxation speed down to low energy conformations is significantly enhanced allowing application to larger systems or evaluation of more starting onformations compared to regular MD. Combined with a generalized Born implicit solvation model the low energy cyclic peptide conformations show also reasonable agreement with experiment. Applications of the approach to the prediction of protein loop regions to improve protein modelling based on sequence similarity to proteins with known structure will also be presented.