We show that the interplay between excluded volume effects, hydrophobicity, and hydrogen bonding of a tube-like representation of a polypeptide chain gives rise to free energy landscapes that exhibit a small number of metastable minima corresponding to common structural motifs observed in proteins. The complexity of the landscape increases only moderately with the length of the chain. Analysis of the temperature dependence of these landscapes reveals that the stability of specific metastable states is maximal at a temperature close to the mid-point of folding. These mestastable states are therefore likely to be of particular significance in determining the generic tendency of proteins to aggregate into potentially pathogenic agents.PACS numbers: 87.15.Aa, 87.14.EeThe mechanism by which proteins fold reliably into their native states is frequently described by using the concept of a free energy landscape (FEL) [1,2,3,4]. Under physiological conditions, the region of configuration space associated with the native state has the lowest free energy and is therefore thermodynamically the most stable. In addition to the native and fully unfolded states, intermediate structures have been detected in the folding and misfolding processes of many proteins [5,6,7,8]. These metastable states can play an important role in the folding process, or be kinetic traps that interfere with correct folding. The experimental characterization of these states remains challenging, as they are often transient or disordered, but significant progress in this direction has recently been made by combining experiment with theory [8,9,10]. Since specific metastable states can increase the probability of misfolding and aggregation [11,12], it is important to understand the mechanism of their formation. Indeed, a global characterization of the FEL is crucial to a full understanding of the relationship between the native and metastable states, and the interplay between folding and misfolding.A complete characterization of the FEL by computational methods requires that the free energies of the native and all non-native structures be calculated. Two problems arise in this exercise: the technical issue of sampling a vast number of configurations separated by a variety of free energy barriers, and the conceptual difficulty of choosing an appropriate set of coordinates to describe the FEL [13]. Sampling is typically performed in the vicinity of an ensemble of folding or unfolding pathways, and the resulting free energy is then projected on to a subspace defined by one or two order parameters [3]. For the resulting surfaces to provide insight into thermodynamics and dynamics, it is crucial that the chosen order parameters are able to detect the relevant details of the FEL. In cases where a long dynamic trajectory that explores a large region of configuration space is available, it is possible to mitigate the order parameter problem by using the dynamics to cluster the configurations and disconnectivity graphs to represent the results [14]. To maintain th...