A parallel searching algorithm based on eigenvector-following is used to generate databases of minima and transition states for an all-atom model of the peptide Ac(ala)(3)NHMe and for a simplified bead model of a protein. We analyze the energy landscapes of both systems using disconnectivity graphs based upon both potential energy and free energy. This approach highlights the role of vibrational entropy in determining the relative free energy of local minima. Thermodynamic properties for Ac(ala)(3)NHMe calculated using the superposition approach are in reasonable agreement with parallel-tempering Monte Carlo simulations. (C) 2003 American Institute of Physics.