Protein structure prediction using global optimization by basin-hopping with NMR shift restraints

Enhancing Biomolecular Simulations with Hybrid Potentials Incorporating NMR Data

Some recent advances in biomolecular simulation and global optimization have used hybrid restraint potentials, where harmonic restraints that penalize conformations inconsistent with experimental data are combined with molecular mechanics force fields. These hybrid potentials can be used to improve the performance of molecular dynamics, structure prediction, energy landscape sampling, and other computational methods that rely on the accuracy of the underlying force field. Here, we develop a hybrid restraint potential based on NapShift, an artificial neural network trained to predict protein nuclear magnetic resonance (NMR) chemical shifts from sequence and structure. In addition to providing accurate predictions of experimental chemical shifts, NapShift is fully differentiable with respect to atomic coordinates, which allows us to use it for structural refinement. By employing NapShift to predict chemical shifts from the protein conformation at each simulation step, we can compute an energy penalty and the corresponding hybrid restraint forces based on the difference between the predicted values and the experimental chemical shifts. The performance of the hybrid restraint potential was benchmarked using both basin-hopping global optimization and molecular dynamics simulations. In each case, the NapShift hybrid potential improved the accuracy, leading to better structure prediction via basin-hopping and increased local stability in molecular dynamics simulations. Our results suggest that neural network hybrid potentials based on NMR observables can enhance a broad range of molecular simulation methods, and the prediction accuracy will improve as more experimental training data become available.

The membrane-active amphibian peptide caerin 1.8 inhibits fibril formation of amyloid β1-42

The amphibian host-defense peptide caerin 1.8 [(1)GLFKVLGSV(10)AKHLLPHVVP(20)VIAEKL(NH2)] inhibits fibril formation of amyloid β 1-42 [(1)DAEFRHDSG(10)YEVHHQKLVF(20)FAEDVGSNKG(30)AIIGLMVGGV(40)VIA] [Aβ42] (the major precursor of the extracellular fibrillar deposits of Alzheimer's disease). Some truncated forms of caerin 1.8 also inhibit fibril formation of Aβ42. For example, caerin 1.8 (1-13) [(1)GLFKVLGSV(10)AKHL(NH2) and caerin 1.8 (22-25) [KVLGSV(10)AKHLLPHVVP(20)VIAEKL(NH2)] show 85% and 75% respectively of the inhibition activity of the parent caerin 1.8. The synthetic peptide KLVFFKKKKKK is a known inhibitor of Aβ42 fibril formation, and was used as a standard in this study. Caerin 1.8 is the more effective fibril inhibitor. IC50 values (± 15%) are caerin 1.8 (75 μM) and KLVFFKKKKKK (370 μM). MALDI mass spectrometry shows the presence of a small peak corresponding to a protonated 1:1 adduct [caerin 1.8/Aβ42]H(+). Molecular dynamics simulation suggests that both hydrogen bonding and hydrophobic interactions between Aβ42 and caerin 1.8 facilitate the formation of a 1:1 complex in water. Fibril formation from Aβ42 has been proposed to be based around the (16)KLVF(20)F region of Aβ42; this region in the 1:1 complex is partially blocked from attachment of a further molecule of Aβ42.

Energy landscapes of a hairpin peptide including NMR chemical shift restraints

Methods recently introduced to improve the efficiency of protein structure prediction simulations by adding a restraint potential to a molecular mechanics force field introduce additional input parameters that can affect the performance. Here we investigate the changes in the energy landscape as the relative weight of the two contributions, force field and restraint potential, is systematically altered, for restraint functions constructed from calculated nuclear magnetic resonance chemical shifts. Benchmarking calculations were performed on a 12-residue peptide, tryptophan zipper 1, which features both secondary structure (a β-hairpin) and specific packing of tryptophan sidechains. Basin-hopping global optimization was performed to assess the efficiency with which lowest-energy structures are located, and the discrete path sampling approach was employed to survey the energy landscapes between unfolded and folded structures. We find that inclusion of the chemical shift restraints improves the efficiency of structure prediction because the energy landscape becomes more funnelled and the proportion of local minima classified as native increases. However, the funnelling nature of the landscape is reduced as the relative contribution of the chemical shift restraint potential is increased past an optimal value.

Protein structure prediction: assembly of secondary structure elements by basin-hopping

The prediction of protein tertiary structure from primary structure remains a challenging task. One possible approach to this problem is the application of basin-hopping global optimization combined with an all-atom force field. In this work, the efficiency of basin-hopping is improved by introducing an approach that derives tertiary structures from the secondary structure assignments of individual residues. This approach is termed secondary-to-tertiary basin-hopping and benchmarked for three miniproteins: trpzip, trp-cage and ER-10. For each of the three miniproteins, the secondary-to-tertiary basin-hopping approach successfully and reliably predicts their three-dimensional structure. When it is applied to larger proteins, correctly folded structures are obtained. It can be concluded that the assembly of secondary structure elements using basin-hopping is a promising tool for de novo protein structure prediction.