Predicting the aggregation propensity of mAb formulations from molecular dynamics simulations

点击翻译按钮获取中文摘要

Context & impact: Aggregation is increasingly thought to occur through the partial unfolding of protein structure to expose sites that are more prone to self-interaction. Overall, the self-association of proteins is influenced by a combination of surface properties that determine colloidal stability and propensity for surface interactions, the extent and kinetics of global and local unfolding, and the solvent accessibility and aggregation-propensity of local sequences. All of these are modulated by the physical environment provided by the formulation, which can thus alter both the kinetics and dominant pathways of aggregation. Identifying the specific influence of formulations on aggregation kinetics and mechanism for a give protein, requires considerable experimental characterisation, while few generalities can be reliably used in the design of formulations for new proteins of interest. There has been considerable recent growth of experimental characterisation of protein aggregation mechanisms in a range of formulations, and an increased role of computational molecular dynamics simulations to provide insights into the molecular events that lead to aggregation. Building on this, there is now significant potential for computational approaches to begin to predict the impact of formulations on protein stability and aggregation.Aims and objectives: The aim of this project will be to develop a workflow of modelling and simulation approaches, that can provide molecular-level insights into the experimental aggregation behaviour of mAbs under a range of conditions, and provide a basis for their use in prediction of formulation stability.Research methodology:The project will collaborate with CSL to explore the use of all-atom and course-grain (CG) molecular dynamics simulations of mAbs under formulation conditions, and in contact with surfaces, to investigate ability of simulations to identify mAb behaviours (protein-protein and protein-surface interactions) that correlate to known stability characteristics. Statistical approaches will be used to correlate simulation features (e.g. RMSD, RMSF, APR exposure) to known experimental properties, to elucidate potential aggregation initiation mechanisms. A. hybrid approach will also be explored in which all-atom simulations are clustered to provide alternative conformations for CG simulations. Thus, the project will also train the student in the latest digital skills, including machine learning. It is anticipated that the findings and approaches could be used to predict (& understand) historical CSL mAb datasets. Anticipated outcomes: Insights into the predictive & mechanistic power of molecular dynamics in mAb formulations. A road map for further MDS platform development & future implementation into CSL workflows. Several peer-reviewed publications. The project is aligned directly to the EPSRC Manufacturing the Future theme.