Membrane proteins are responsible for the communication between cells and their environments. They are indispensable to the expression of life phenomena and also implicated in a number of diseases. Nevertheless, the studies on membrane proteins are far behind those on water-soluble proteins, primarily due to their low structural stability. Introduction of mutations can enhance their thermostability and stability in detergents, but the stabilizing mutations are currently identified by experiments. The recently reported computational methods suffer such drawbacks as the exploration of only limited mutational space and the empiricism whose results are difficult to physically interpret. Here we develop a rapid method that allows us to treat all of the possible mutations. It employs a free-energy function (FEF) that takes into account the translational entropy of hydrocarbon groups within the lipid bilayer as well as the protein intramolecular hydrogen bonding. The method is illustrated for the adenosine A(2a) receptor whose wild-type structure is known and utilized. We propose a reliable strategy of finding key residues to be mutated and selecting their mutations, which will lead to considerably higher stability. Representative single mutants predicted to be stabilizing or destabilizing were experimentally examined and the success rate was found to be remarkably high. The melting temperature T-m for two of them was substantially higher than that of the wild type. A double mutant with even higher T-m was also obtained. Our FEF captures the essential physics of the stability changes upon mutations.