Prediction of protein loop structures using a local move Monte Carlo approach and a grid-based force field

Meng Cui, Mihaly Mezei and Roman Osman

Department of Physiology and Biophysics
Mount Sinai School of Medicine, NYU
New York, NY 10019

Published in Protein Eng Des Sel, Vol. 21(12) pl 729-735 (1008).

Overall strategy

Two types of MC moves are performed:
  • simple moves for torsion angles of amino acid side chains in the loop; and
  • local moves for the seven consecutive backbone torsion angles in a window of loops.

    To reduce the computational cost of energy evaluation in this method we developed a potential map, which include van der Waals, electrostatics, hydrophobic as well as hydrogen bond potentials to replace the bulk protein environment and solvation effect.

    We reparametrized the distance-dependent dielectric function of Mehler and Solmayer to work in conjunction with the desolvation and hydrogen-bonding terms used.

    To improve the efficiency of LMMC sampling we employed a two-step protocol:

  • Loop sampling at high temperature by using a reduced potential map to lower the energy barriers for exploring a larger conformational space;
  • Clustering of loop conformations extracted from the high-temperature run followed by multiple simulated annealing simulations to identify the loop conformations in local energy minima.

    Methods

    Local Move Monte Carlo Sampling: LMMC loop sampling starts with changing one backbone torsion angle followed by the adjustment of the six subsequent torsions to allow the rest of the loop to remain in its original position while preserving all bond lengths and bond angles.

    Local move

    Force field

    RESULTS

    Loop structures of 5cpa (231-237) produced by local move MC method at 5000K and followed by clustering to generate 100 representative conformations. Black stick represents the crystal loop structure, and gray wires represent the 100 representative loop conformations.

    Total energies vs. RMSD of the 100 representative loop conformations of 5cpa (231-237) from local move MC sampling (before simulated annealing). The RMSD calculations are based on the backbone atoms of the crystal and representative loop structures. Total energies vs. RMSD (100 predicted loop conformations (the energies higher than 27900 kcal/mol are not shown) of 8tln (248-255) from local move MC simulated annealing). The RMSD calculations are based on the backbone atoms of the crystal and predicted loop structures

    Crystal packing effects

    Protein (3grs) forms trimer in the crystal packing. The loops to be predicted are located at the interface of two monomers. The gray wires represent the protein structures in trimer form; black sticks represent loops (L1, L2 and L3 (83-89) are from different monomers) to be predicted.
    The crystal, initial and predicted loop structures in the protein dimer of 8tln (248-255). The gray wires represent the crystal structures, the gray stick represents the initial loop structure (6.0 Å RMSD from crystal loop structure), and the black stick represents the MC predicted loop structure (0.67 Å RMSD from crystal loop structure).

    Comparison with other methods

    Comparison of loop prediction results from local move MC (LMMC) simulated annealing method and others (The results are based on the backbone global RMSD (in Å) between the crystal and predicted loop structures).
    *: Crystal packing information is included for loop prediction by LMMC method only. (a). Results from our Local Move Monte Carlo approach; (b). Results from our Local Move Monte Carlo approach followed by energy Minimization; (c). Data taken from Table 8 of Vlijmen and Karplus; (d). Data taken from Table 9 of Deane and Blundell; (e). Data taken from Table VI of Deane and Blundell; (f). Data taken from Table 2 of Fiser et al.; (g). Data taken from Table II of Michalsky et al.; (h). Data taken from Table I of Rohl et al.

    The calculations were performed with the MMC Monte Carlo program, available at http://inka.mssm.edu/~mezei/mmc

    Back to the Mezei Lab home page

    Last modified: 03/20/2009 (MM)