Pathological folding and aggregation of the amyloid beta-protein (Abeta) are widely perceived as central to understanding Alzheimer's disease (AD) at the molecular level. Experimental approaches to study Abeta self-assembly are limited, because most relevant aggregates are quasi-stable and inhomogeneous. In contrast, simulations can provide significant insights into the problem, including specific sites in the molecule that would be attractive for drug targeting and details of the assembly pathways leading to the production of toxic assemblies. Here we review computer simulation approaches to understanding the structural biology of Abeta. We discuss the ways in which these simulations help guide experimental work, and in turn, how experimental results guide the development of theoretical and simulation approaches that may be of general utility in understanding pathologic protein folding and assembly.