A knowledge-based potential for a rotamer library was developed to design p
rotein sequences. Protein side-chain conformations are represented by 56 te
mplates. Each of their fitness to a given structural site-environment is ev
aluated by a combined function of the three knowledge-based terms, i.e. two
-body side-chain packing, one-body hydration and local conformation. The nu
mber of matches between the native sequence and the structural site-environ
ment in the database and that of the virtually settled mismatches, counted
in advance, were transformed into the energy scores. In the best-14 test (a
ssessment for the reproduction ability of the native rotamer on its structu
ral site within a quarter of 56 fitness rank positions), the structural sta
bility analysis on mutants of human and T4 lysozymes and the inverse-foldin
g search by a structure profile against the sequence database, this functio
n performs better than the function deduced with the conventional normaliza
tion and our previously developed function. Targeting various structural mo
tifs, de novo sequence design was conducted with the function. The sequence
s thus obtained exhibit reasonable molecular masses and hydrophobic/hydroph
ilic patterns similar to the native sequences of the target and act as if t
hey were the homologs to the target proteins in BLASTP search. This signifi
cant improvement is discussed in terms of the reference state for normaliza
tion and the crucial role of short-range repulsion to prohibit residue bump
s.