A Multiscale Model for Quantitative Prediction of Insulin Aggregation Nucleation Kinetics

Abstract

We combined kinetic, thermodynamic, and structural information from single molecule (protein folding) and two molecule (protein association) explicit-solvent simulations for determination of kinetic parameters in protein aggregation nucleation with insulin as model protein. A structural bioinformatics approach was developed to account for heterogeneity of aggregation-prone species with the transition complex theory, developed for native protein-receptor interactions, found applicable in modeling association kinetics involving this non-native species. We show that a key simplification arises from presence of only a few relevant modes for non-native association kinetics and that it is necessary to explicitly account for conformational rearrangement of a diffusional intermediate leading to the formation of aggregation pathway dimer and small oligomers. The kinetic parameters thus obtained were used in a population balance model and very accurate predictions for aggregation nucleation time varying over two orders of magnitude with changes in concentration of insulin or an aggregation-inhibitor ligand were obtained while an empirical parameter set was not found to be transferable for prediction of ligand effects. This physically determined kinetic parameter set also provided several insights into the mechanism of aggregation nucleation. Finally we discuss a route for application of our approach in high-throughput computational screening of ligands for inhibiting aggregation.