Static and dynamic properties of concentrated monoclonal antibody solutions
Title: |
Static and dynamic properties of concentrated monoclonal antibody solutions |
DNr: |
NAISS 2025/5-315 |
Project Type: |
NAISS Medium Compute |
Principal Investigator: |
Peter Schurtenberger <peter.schurtenberger@fkem1.lu.se> |
Affiliation: |
Lunds universitet |
Duration: |
2025-06-01 – 2026-06-01 |
Classification: |
10402 |
Keywords: |
|
Abstract
Antibodies have emerged as essential tools in modern medicine. Although antibodies share a common structural architecture characterized by an anisotropic Y shape, their exceptional versatility arises primarily from the significant variability in their amino acid composition. This intrinsic flexibility enables the precise engineering of biomolecules tailored to specific therapeutic goals and diverse biomedical applications. However, subtle structural variations among antibodies profoundly influence their behavior in solution, altering both static and collective dynamic properties. Currently, predicting how specific amino acid sequences and the arising charge distribution affect the physical properties of antibody solutions remains challenging. This gap underscores the necessity of developing robust and generalizable methodologies capable of linking molecular-level structural features to macroscopic solution behaviors.
In our previous project (NAISS 2024/5-292), we took an important step toward this goal by demonstrating that a bead-based computational model incorporating heterogeneous charge distributions effectively captures both the structure factor and viscosity across a range of antibody solution concentrations and ionic strengths.
Our continuation proposal seeks to further elucidate how variations in antibody internal structure translate into measurable changes in solution properties, and how these changes can be treated through an appropriate coarse-graining strategy. We will use MD simulations, also complemented with machine learning approaches, starting from four distinct antibody structures for which we have extensive experimental data available. These structures differ fundamentally: while two exhibit a large overall positive charge and a relatively homogeneous charge distribution, the other two feature well-defined oppositely charged patches that give rise to additional attractive contributions to the protein-protein interactions. The available experimental data demonstrate the enormous impact of the different structures on the relevant solution properties, e.g. on the relative viscosity and the occurrence of liquid-liquid phase separation at high concentrations.
The core part of our research will systematically explore the effects of structural modifications on solution behavior by examining variations in i) internal charge distribution, ii) spatial localization of attractive patches, iii) overall antibody charge, and iv) short-range van der Waals and hydrophobic interactions. We will generate different coarse-grained models with different structural features and run MD simulations of multiple antibodies that explicitly include Coulomb electrostatic interactions and salt ions at different antibody concentrations, preserving the anisotropic antibody architecture. From these we will calculate the structure factors and dynamical properties such as the viscosity. The simulation-generated data will also form extensive datasets that we will utilize to train neural networks. These ML models will then pinpoint the structural characteristics that most accurately reproduce experimentally observed solution behaviors for the antibodies under investigation. The use of ML applied to coarse-grained models constitutes a significant methodological advancement over current techniques, which typically focus on atomistic-level investigations of individual antibodies.
In parallel, our investigation will also provide insight on how structural variations influence the effective interaction potential (potential of mean force, PMF) between antibodies, which is directly linked to experimentally measurable parameters such as the second virial coefficient B2 or the diffusion parameter kD.