The robustness of a system is a property that pervades all aspects of Nature. The ability of a system to adapt itself to perturbations due to internal and external agents, to aging, to wear, to environmental changes is one of the driving forces of evolution. At the molecular level, understanding the robustness of a protein has a great impact on the in silicon design of polypeptide chains and drugs; the chance of computationally checking the ability of a protein to preserve its structure and function in the native state can lead to the design of new compounds that can work in a living cell more effectively. Inspired by the well known robustness analysis framework used in Electronic Design Automation, we introduced a notion of robustness for proteins and two dimensionless quantities: the energetic yield and the energetic relative entropy. We used the energetic yield in order to quantify the robustness of a protein, and to detect sensitive regions and sensitive residues in the protein, whereas we adopted the energetic relative entropy to measure the discrepancy between two potential energy distributions. Subsequently, we implemented a new robustness-centred protein design algorithm called Robust-Protein-Design (RPD); the aim of the algorithm is to discover new conformations with a specific function with high yield values. We performed an extensive characterization of the robustness property of many peptides, proteins, and drugs. Moreover, we found that robustness and relative entropy are conflicting objectives which constitute a trade-off useful as design principle for new proteins and drugs. Finally, we used the RPD algorithm on the Crambin protein (1CRN): the obtained results confirm that the algorithm was able to find out a Crambin-like protein that is 23% more robust than the wild type. (c) 2010 Elsevier Ltd. All rights reserved.