UC Berkeley

My doctoral research involves the characterization of the structure, kinetics, and function of amyloid-β (Aβ) proteins by computational means via atomistic and coarse-grained molecular dynamics simulations. Aβ has critical clinical relevance as one of the key hallmarks of Alzheimer’s disease pathology. My research has four primary foci. The first of these is studying the properties of the Aβ monomer, an intrinsically disordered protein (IDP), using all-atom simulations. Using a combination of two enhanced sampling techniques - replica-exchange molecular dynamics simulations (REMD) and temperature cool walking (TCW), I have shown that the addition of paramagnetic tags in paramagnetic relaxation enhancement (PRE) experiments of IDPs, perturbs the structural ensembles of the Aβ monomer with an increase in structural order, and the PRE experimental observables are thus not a true representation of the unmodified monomers in solution. Very few experimental techniques can provide residue-specific structural information about IDPs because of their disordered nature, and my work provides valuable insights into the usefulness of commonly used PRE experiments in the IDP field.

IDP structural ensembles are usually generated using REMD simulations with fixed charge protein models. Most computationally generated IDP ensembles using physics-based models are more ordered and compact than expected. The current focus is mostly on improving or modifying parameters in fixed charge force fields to generate more accurate conformational ensembles for IDPs. By comparing different sampling techniques and fixed charge force fields (with and without IDP-specific parameter modification) for the Aβ42 and Aβ43 peptide, I show that the sampling method used to generate the ensemble is equally important as the “correct” force field. The IDP ensembles generated using TCW have better convergence and experimental agreement than the REMD-ensemble for same amount of sampling. Thus, TCW is a better sampling alternative to REMD simulations.

Fixed charge force fields used in IDP simulations are parameterized on folded protein data, and thus predict overly structured and globular configurations for IDPs, which are usually more disordered and solvent exposed compared to folded proteins. Consequently, it is important that the molecular interactions are modeled as accurately as possible during IDP simulations. In chapter 4, using the cationic 24-residue Histatin 5 peptide as a test system, I show that the computationally generated ensemble using the polarizable AMOEBA force field is more consistent with experimental radius of gyration and secondary structure data. Thus, the many-body polarization effect that is ignored in fixed charge force field is important for simulating IDP systems across a range of solvent-exposed to folded states, capturing the true breadth of structural biology.

The last major emphasis of this dissertation research, is investigating of kinetic elongation mechanisms of amyloid fibril (aggregates of Aβ monomer) using an in-house coarse-grained protein model. In chapter 5, I studied the mechanism of amyloid fibrils elongation via binding of monomers from solution, and demonstrated that the monomer structure only influences the kinetics, but not the overall binding mechanism. This result provides a fundamental understanding of growth of amyloid fibrils.