“…The nature of these problems reinforces the point that the scoring function must be fast and accurate enough for good scoring performance, which provides motivation, challenge, and validation to the physics-based energy functions (e.g., modern molecular mechanics plus continuum dielectric solvent models). While there exists some evidence that a GB model might not help for this purpose (Morozov et al, 2003), a number of other studies have shown that the physics-based energy functions perform well in distinguishing native and near-native folds (Dominy and Brooks, 2001;Feig and Brooks, 2002;Felts et al, 2002;Zhu et al, 2003) and loop conformations (Fiser et al, 2002;Forrest and Woolf, 2003) from a large set of nonnative decoys. Recent examination of GB and PB implicit solvation models in the context of computational protein design revealed a systematic bias that the burial of polar amino acids in the protein interior is more favored than that of nonpolar ones, which can be problematic in protein design (Jaramillo and Wodak, 2005).…”