Amyloid protein aggregation characterizes many neurodegenerative disorders, including Alzheimer's, Parkinson's, and Creutzfeldt-Jakob disease. Evidence suggests that amyloid aggregates may share similar aggregation pathways, implying simulation of full-length amyloid proteins is not necessary for understanding amyloid formation. In this study, we simulate GNNQQNY, the N-terminal prion-determining domain of the yeast protein Sup35 to investigate the thermodynamics of structural transitions during aggregation. Utilizing a coarse-grained model permits equilibration on relevant time scales. Replica-exchange molecular dynamics is used to gather simulation statistics at multiple temperatures and clear energy traps that would aversely impact results. Investigating the association of 3-, 6-, and 12-chain GNNQQNY systems by calculating thermodynamic quantities and orientational order parameters, we determine the aggregation pathway by studying aggregation states of GNNQQNY. We find that the aggregation of the hydrophilic GNNQQNY sequence is mainly driven by H-bond formation, leading to the formation of β-sheets from the very beginning of the assembly process. Condensation (aggregation) and ordering take place simultaneously, which is underpinned by the occurrence of a single heat capacity peak. Proteins 2013; 81:1141–1155. © 2013 Wiley Periodicals, Inc.