Conformational search is a process of finding low-energy conformations of molecular systems by varying user-specified dihedral angles. The method involved variation of dihedral angles to generate new structures and then energy minimizing each of these angles. Low-energy unique conformations are stored while high-energy duplicate structures are discarded. Because molecular flexibility is usually due to rotation of unhindered bond dihedral with little change in bond lengths or bond angles, only dihedral angles are considered in the conformational search. Its goal is to determine the global minimum of the potential energy surface of a molecular system. Several approaches have been applied to the problem of determining low-energy conformations of molecules (Howard and Kollman, 1988). These approaches generally consist of the following steps with differences in details:
1. Selection of an initial structure: The initial structure is the most recently accepted conformation (e.g., energy minimized structure) and remains unchanged during the search. This is often referred to as a random walk scheme in Monte Carlo searches. It is based on the observation that low-energy conformations tend to be similar, therefore starting from an accepted conformation tends to keep the search in a low-energy region of the potential surface. An alternative method, called the usage-directed method, seeks to uniformly sample a low-energy region by going through all previously accepted conformations while selecting each initial structures (Chang et al., 1989). Comparative studies have found the usage-directed scheme to be superior for quickly finding low-energy conformations.
2. Modification of the initial structure by varying geometric parameters: The variations can be either systematic or random. Systematic variations can search the conformational space exhaustively for low-energy conformations. However, the number of variations becomes prohibitive except for the simplest systems. One approach to reduce the dimensionality of systematic variation is to first exhaust variations at a low resolution, then exhaust the new variations allowed by successively doubling the resolution. Random variations choose a new value for one or more geometric parameters from a continuous range or from sets of discrete values. To reduce the number of recurring conformations, several random variations have some sort of quick comparison with the sets of previous structures prior to performing energy minimization of the new structure.
3. Geometry optimization of the modified structure to energy-minimized conformations: The structures generated by variations in dihedral angles are energy-minimized to find a local minimum on the potential surface. Although the choice of optimizer (minimizer) has a minor effect on the conformational search, it is preferable to employ an optimizer that converges quickly to a local minimum without crossing barrier on the potential surface.
4. Comparison of the conformation with those found previously: The conformation is accepted if it is unique and its energy satisfies a criterion. Two types of criteria are used to decide an acceptance of the conformation. Firstly, geometric comparisons are made with previously accepted conformations to avoid duplication. Conformations are often compared by the maximum deviation of torsions or RMS deviation for internal coordinates, interatomic distances, or least-squares superposition of conformers. Because geometry optimization can invert chiral centers, the chiral centers of the modified structures should be checked after the energy minimization. Secondly, the energetic test for accepting a new conformer may be carried out by a simple cutoff relative to the best energy found so far or a Metropolis criterion where higher-energy structures are accepted with a probability determined by the energy difference and a temperature, for example, exp(- AE/kT).
Was this article helpful?