Aggregation of intrinsically disordered proteins (IDPs), such as amyloid beta peptide, can cause serious health problems, associated with disorders including Alzheimer disease. Due to the lack of a stable structure and transient nature, such proteins and peptides are often very difficult or even impossible to study using experimental approaches. Therefore, usage of computational tools can provide valuable insight into their dynamics, structural changes, and mechanism of aggregation. Because current force fields were designed to work well for standard proteins with a well-defined native structure and high conformational stability, we examined three force fields most frequently used for studies of proteins, and two variants modified for better performance for IDPs on an example of monomeric amyloid beta 42 (Aβ42) with two sampling approaches: single 10 µs long conventional molecular dynamics (CMD) trajectories and 48-replica runs using the replica exchange MD (REMD). We found that newer force fields (Amber FF14SB and CHARMM36m) provided better results than their older versions (Amber FF99SB and CHARMM36), while the specially modified version for the IDPs (FF14SB_IDPs) yielded similar results to its parent, improving sampling using CMD simulations, hence allowing to achieve a similar level of accuracy at significantly lower computational costs. With sufficient sampling, the newer force fields provided good agreement with the available experimental data. We also analyzed the physical basis of different behaviors of force fields and sampling methods, concluding that in CHARMM interactions with water play a much more important role than in Amber force fields. This explains why, in CHARMM force fields, the monomeric Aβ42 is less stable and more hydrophilic, having a greater solvent accessible surface area.
Read full abstract