This dissertation is composed of two parts. The first part includes Chapters 1-4, which focuses on molecular dynamics (MD) studies of amyloid-β (Aβ) aggregation. Our results reveal that as in early stages of aggregation as monomer, there is a conformation, namely strand-loop-strand (SLS), which may be active for Aβ aggregation. Our simulations show that the SLS is stable at various conditions. Also, the topology of the SLS resembles that of Aβ conformation in aggregates. We also find that a well-known aggregation-disrupting mutation, F19T, destabilizes the SLS, which supports our idea that the SLS is aggregation-active. By simulating amyloid-like Aβ oligomers, we found that dehydration of charged side chains such as E22, D23 and K28 is highly unfavorable despite that D23 and K28 are fou...[ Read more ]

This dissertation is composed of two parts. The first part includes Chapters 1-4, which focuses on molecular dynamics (MD) studies of amyloid-β (Aβ) aggregation. Our results reveal that as in early stages of aggregation as monomer, there is a conformation, namely strand-loop-strand (SLS), which may be active for Aβ aggregation. Our simulations show that the SLS is stable at various conditions. Also, the topology of the SLS resembles that of Aβ conformation in aggregates. We also find that a well-known aggregation-disrupting mutation, F19T, destabilizes the SLS, which supports our idea that the SLS is aggregation-active. By simulating amyloid-like Aβ oligomers, we found that dehydration of charged side chains such as E22, D23 and K28 is highly unfavorable despite that D23 and K28 are found to form buried salt bridges in aggregates. We thus suggest that the dehydration may present a barrier for aggregation. Aggregation-prone mutations, such as E22 and D23, may bypass this barrier without the dehydration. The second part covers Chapters 5-6, where an effort in developing a coarse-grained protein model is presented. Our model is based on a developed CG water model. By parameterizing potential functions through solvation properties of organic compounds and conformational properties of peptides, our model is able to reproduce solvation free energies of various organic compounds, transfer free energy of amino acids from non-polar to polar solvent, and the helical propensities of Gly, Ala, Leu and Val with a high accuracy. It can fold polypeptides into various secondary structures without any bias potential, and recognize the preference for different secondary structures by different peptide sequences. It is of comparable accuracy but is about 1000~2000 times faster than current all-atom models. Thus, it is promising for the study of protein folding upon its full development.