Dipartimento di Scienze Farmaceutiche − Sezione di Chimica Generale e Organica “Alessandro Marchesini”, Università degli Studi di Milano, Via Venezian, 21, 20133 Milano, Italy
Auto-generated profile, based on publicly available information
Andrea Bazzoli
Dipartimento di Scienze Farmaceutiche − Sezione di Chimica Generale e Organica “Alessandro Marchesini”, Università degli Studi di Milano, Via Venezian, 21, 20133 Milano, Italy
Accelerated molecular dynamics (aMD) protocols were assessed on predicting the secondary structure of eight peptides, of which two are helical, three are β-hairpins, and three are disordered. Protocols consisted of combinations of three force fields (ff99SB, ff14SB, ff19SB) and two explicit solvation models (TIP3P and OPC), and were evaluated in two independent aMD simulations, one starting from an extended conformation, the other starting from a misfolded conformation. The results of these analyses indicate that all three combinations performed well on helical peptides. As for β-hairpins, ff19SB performed well with both solvation methods, with a slight preference for the TIP3P solvation model, even though performance was dependent on both peptide sequence and initial conformation. The ff19SB/OPC combination had the best performance on intrinsically disordered peptides. In general, ff14SB/TIP3P suffered the strongest helical bias.
One of the strategies of peptide–protein docking is to pregenerate an ensemble of peptide conformations in the absence of the receptor, and then dock them as rigid bodies onto its surface. Success of this strategy requires that the scoring function that drives the pregeneration step be able to discriminate in favor of conformations that resemble the native bound conformation. Here we present a study on the discrimination of peptide native bound conformations as achieved without receptor by the “cen_std+score4L” Rosetta energy function, a low-resolution scoring function equivalent to one chosen for other tasks where the modeling of solvent effects is of special importance. The cen_std+score4L function was able to assign, on average, lower energies to native-like than to non-native decoy conformations for only 3 of our 18 test peptides; it also ranked one or more native-like decoys in the top 1% for only 2 peptides. However, by optimizing the weights of the energy terms that define the cen_std+score4L function, native discrimination improved substantially: Native-like decoys were assigned lower energies than non-native decoys for 16 peptides, with a discrimination signal larger than noise for 9 peptides, that is, 50% of the test set. And for 9 peptides, too, native-like decoys ranked in the top 1%. An ensuing energetic analysis of native-like versus non-native decoys suggests that native peptide conformations have solvation and non-local electrostatics that poorly recapitulate those of native protein conformations. Native peptide conformations are also characterized by few backbone–backbone H-bonds and by lack of compactness, presumably to optimize interaction with the receptor. Overall, this study lays groundwork for pregenerating dockable peptide conformations with Rosetta, whether the subsequent docking will be performed by Rosetta or some other software.