We propose an automated protocol for designing the energy landscape
suitable for the description of a given set of protein sequences
with known structures, by optimizing the parameters of the energy
function. The parameters are optimized so that not only the global
minimum-energy conformation becomes native-like, but also the
conformations distinct from the native structure have higher
energies than those close to the native one, for each protein
sequence in the set. In order to achieve this goal, one has to
sample protein conformations that are local minima of the energy
function for given parameters. Then the parameters are optimized
using linear approximation, and then local minimum conformations are
searched with the new energy parameters. We develop an algorithm
that repeats this process of parameter optimization based on linear
approximation, and conformational optimization for the current
parameters, ultimately leading to the optimization of the energy
parameters. We test the feasibility of this algorithm by optimizing
a coarse grained energy function, called the UNRES energy function,
for a set of ten proteins.