Understanding the initial steps of protein aggregation leading to the formation of amyloid fibrils remains a challenge. Here, the kinetics of such a process is determined for a misfolding protein model, ADA2h. The double nature of the very early kinetics suggests a step model of aggregation, where the denatured polypeptide folds into an aggregated beta-intermediate that subsequently reorganises into a more organised beta-sheet-richer structure that finally results in amyloid fibre formation. To determine the regions of the protein involved in amyloidosis, we have analysed a series of mutants previously made to study ADA2h folding. Using the algorithm TANGO, we have designed mutants that should enhance or decrease aggregation. Experimental analysis of the mutants shows that the C terminus of the molecule (comprising the last and edge beta-strand) is the major contributor to amyloid fibril formation, in good agreement with theoretical predictions. Comparison with proteins with similar topology reveals that family folds do not necessarily share the same principles of protein folding and/or aggregation.