Overlap distribution in random and designed heteropolymers

We study the overlap between low-energy states in lattice models of heteropolymers with contact interactions. The overlap distribution gives information on the degree of correlation in the energy landscape. Designed sequences have rather correlated energy landscapes, which favor fast folding kinetics. Chains with random interactions have much less correlated energy landscapes. It is indeed believed that the mean-field theory for this model coincides with the Random Energy Model, whose different low-energy states are completely unrelated. This picture has been supported by numerical studies of maximally compact configurations. Without applying this constraint, we find that the overlap distribution is indeed bimodal as expected, but it has a broad peak at large overlap, indicating a non-vanishing width for the valleys of low-energy states. This feature probably plays an important role in the kinetics of the model. It is not evident that the range of such correlations shrinks to zero for large systems. The range of the correlations seems to be influenced by the number of contacts per residue in the ground state: the smaller this quantity, the larger the correlations. Received 16 August 2000