Shape-based coarse graining of proteins permits anisotropic intermolecular interactions modulating protein solubility. Together with the ion–protein dispersion, it allows the calculation of theB₂₂coefficients without experimental fitting.

β-lactoglobulin (BLG) forms amyloid-like aggregates at high temperatures, low pH, and low ionic strengths. At a pH below 2, BLG undergoes hydrolysis into peptides, with N-terminal peptides 1–33 and 1–52 being prone to fibrillization, forming amyloid-like fibrils. Due to their good mechanical properties, BLG amyloids demonstrate great potential for diverse applications, including biosensors, nanocomposites, and catalysts. Consequently, further studies are essential to comprehensively understand the factors governing the formation of BLG amyloid-like morphologies. In this study, all-atom molecular dynamics simulations were employed to explore the aggregation of N-terminal 1–33 and 1–52 BLG peptides under conditions of pH 2 and at 10 mM NaCl concentration. The simulations revealed that the peptides spontaneously assembled into aggregates of varying sizes. The aggregation process was enabled by the low charge of peptides and the presence of hydrophobic residues within them. As the peptides associated into aggregates, there was a concurrent increase in β-sheet structures and the establishment of hydrogen bonds, enhancing the stability of the aggregates. Notably, on average, 1–33 peptides formed larger aggregates compared to their 1–52 counterparts, while the latter exhibited a slightly higher content of β-sheets and higher cluster orderliness. The applied approach facilitated insights into the early stages of amyloid-like aggregation and molecular-level insight into the formation of β-sheets, which serve as nucleation points for further fibril growth.

The state of proteins in aqueous solution is determined by weak, nonspecific interactions affected by pH, solvent composition, and ionic strength. Protein–protein interactions play a crucial role in determining protein stability and solubility. The second osmotic coefficient (B₂₂) provides insight into effective interactions between proteins in solution. Models for calculating B₂₂ are valuable for estimating interactions, explaining measured phenomena, and reducing experimental time. However, existing models, like the Derjaguin–Landau–Verwey–Overbeek (DLVO) theory, assume a simple spherical shape for proteins. Owing to the fact that proteins exhibit diverse shapes and charge distributions, influencing their electrostatic properties and overall interactions, DLVO accuracy is significantly reduced for nonspherical proteins. To address this limitation, we introduce the xDLVO-CGhybr model, which combines Poisson–Boltzmann (PB) and Debye–Hückel (DH) theories to account for electrostatic interactions between proteins. PB is used for short intermolecular distances (<2 nm) with an all-atom resolution, while DH is employed for longer distances on a coarse-grained level. Additionally, xDLVO-CGhybr incorporates an improved coarse-grained Lennard-Jones (LJ) potential derived directly from the all-atom potential to capture dispersion interactions. This model improves the calculated B₂₂ values compared to existing models and can be applied to proteins with arbitrary shape and charge under various solvent conditions (up to 1 M monovalent salt concentration). We demonstrate the application of xDLVO-CGhybr to bovine trypsin inhibitor, ribonuclease A, chymotrypsinogen, concanavalin A, bovine serum albumin, and human immunoglobulin type I proteins, validating the model against experimental data.