Abstract: Protein aggregation has been associated with over twenty human diseases, and is a severe problem in the biotechnology industry. This thesis represents an effort to improve our general understanding of aggregation processes, as well as offer suggestions for the abatement of deleterious aggregation. Both computational and experimental approaches are utilized towards this end. Simulations of minimalist model proteins are used to uncover mechanisms of aggregation when proteins fold in close proximity. The simulations were designed such to closely mimic the refolding step common during protein purification from E. coli inclusion bodies. We found that, when considering the amino acid contacts that drive the aggregation process, multiple pathways to aggregation are present, but one is statistically preferred. Additionally, aggregation was shown to occur on two time scales. The first is fast, roughly equivalent to time scale for folding, and occurs between uncollapsed chains. The second is much slower and involves partially---folded intermediates. The sum of these results suggests that, because aggregation may occur through multiple pathways, several methods may be necessary completely to prevent its occurrence. Simulations of lattice-model proteins were used to study the thermodynamics of protein folding and aggregation. Results showed that a protein's free-energy landscape for folding is significantly perturbed by the presence of other proteins. This has dramatic consequences for single-protein conformational preferences: individual molecules unfold at lower temperatures, and compensate by forming inter-protein interactions. Protein-folding mechanisms are substantially different in crowded environments, suggesting that in vitro experiments of protein folding in dilute systems do not necessarily capture protein-folding behavior in vivo. Additionally, for the small systems studied (2-4 proteins), aggregation was entropically driven. Lastly, we showed that native contacts, i.e. contacts that are favored in the protein's native state, are also heavily favored between chains during aggregation. Such behavior has been observed in several experimental studies. The aggregation of protein L was studied in aqueous trifluoroethanol solutions. The aggregation kinetics showed evidence of linear polymerization with a nucleus size of either two or three monomers. Disaggregation kinetics exhibited bi-exponential behavior. Aggregation was not completely reversible over the time scale of the experiments (7days). The remaining aggregates were insoluble, showing that aggregation becomes increasingly difficult to reverse as the size of the aggregates increases.