J 0jo

The PMF can then be calculated as

Ref3 represents a reference state that is independent of solvent exposure of ligand atoms of type i. However, as for pyMi in Refl, the phum are still calculated from reference spheres that are centered at the ligand atoms for each protein-ligand complex. The PMF derived with Ref3 still implicitly consider solvent effects since the density of certain protein atom types is higher at the surface of the protein than inside the protein. However, the degree of which solvation effects are captured is much smaller than in Ref 1.

Other possible reference states, including those that use explicit solvent molecules in the reference sphere, have been discussed for instance by Jerni-gan and Bahar [38]. Reference states such as random mixing [39] that have been developed for protein folding and threading are not applicable here since they have been specifically designed to capture the protein folding energy that is irrelevant for the calculation of protein-ligand binding energies.

Refl and Ref3 are derived 'on the fly' by evaluating the protein atoms occurring around ligand atoms in the protein-ligand complexes used to derive the PMF. In contrast, Ref2 is defined as reference state independent of the protein-ligand interactions. Therefore, the comparison of scoring results obtained with Refl and Ref2 allows us to discuss the issue of a suitable reference state. Refl and Ref3 reflect different degrees of capturing solvation effects in the derived PMF. Therefore, the comparison of scoring results obtained with Refl and Ref3 allows us to assess the effect of the implicit treatment of solvation on the predictive power of the PMF scoring function.

0 0

Post a comment