Definition of reference states

Perhaps the most crucial element of a knowledge-based scoring function is the definition of an appropriate reference state. This reference state must reflect the equilibrium distribution ofprotein atoms around an arbitrary point in the protein ifno interactions with a ligand occur. Unfortunately, the definition of such a reference state is not unique. It is influenced by the presence and treatment of solvent molecules, by ligand atoms, by its size, and by its location. Moreover, the documentation ofknowledge-based scoring functions leaves the reader often in the dark about the exact definition ofthe reference state used. Therefore, we aim here to discuss three viable reference states and compare their effects on the predictive power ofthe PMF scoring function.

Reference state 1

In the above described scoring function the reference state (referred to below as Ref1) is calculated as where ( X complex designates an average over all ligand atoms of type i and over all protein-ligand complexes that are used to derive the PMF. The reference spheres are located around the ligand atoms of type i. Parts of the reference spheres are occupied by the solvent. Therefore, the reference state captures the average solvent exposure of ligand molecules in the protein-ligand complexes. More specifically, it captures the different solvation of types of ligand atoms that are on average more buried in the protein matrix (hydrophobic atoms) or more exposed to the solvent (hydrophilic atoms). Solvation effects are implicitly captured in Ref1 since the different ratios of solvent to protein volume in spherical shells at radius r and in the reference sphere with radius R influence p!i (r) and p tibuit to a different extent (Figure 1). The PMF are calculated by using Equation 2.

0 0

Post a comment