Protein aggregation: Computational modeling and Monte Carlo algorithm development

Abstract: Although organisms have evolved sophisticated cellular mechanisms for regulating their various protein networks, proteins sometimes start to clump together in an uncontrolled way. The aggregation of proteins has been associated with a range of serious diseases. In this thesis we develop and apply theoretical computer models to investigate different aspects of the aggregation process, and its underlying molecular mechanisms. In Papers II and IV we use all-atom Monte Carlo simulations with implicit solvent to characterize two intrinsically disordered proteins known to form fibrillar aggregates: α-synuclein linked with Parkinson’s disease and Aβ linked with Alzheimer’s disease. Paper II investigates the 140-residue α-synuclein in free monomeric form. In Paper IV, we examine the response of Aβ and α-synuclein to a pulling force, and compare with single-molecule experiments. Both studies suggest that fibril-like folds are easily accessible to these proteins. In Paper V the same model is used to investigate the local unfolding dynamics and aggregation propensities of the natively folded SOD1 monomer, which has been associated with the disease amyotrophic lateral sclerosis (ALS). In Paper III we develop a minimal structure-based model for amyloid formation. While most theoretical and computational studies have modeled fibril formation as a 1D growth process, this model allows us to study the interplay between length and width in fibril nucleation. In Papers I and VI we develop and investigate generalized-ensemble techniques for accelerating simulations of systems with rugged free-energy landscapes. We study the flat-histogram method, and an extension of it which assigns higher weight to regions of low diffusivity. A new scheme for estimating diffusion-optimized weights is developed.