Abstract: In this study, computational structure-based protein design was used to increase the thermal stability of wild-type KGF-2. Specifically, the crystal structure of KGF-2 was obtained from the Protein Data Bank (PDB) database. The key residues that may support the structure of KGF-2 and bind to its receptor or ligand were identified to be avoided in the computational protein design. After being truncated with the core region 69−208 aa retained, wild-type KGF-2 was scanned for saturation mutagenesis to determine residues with reduced free energy changes. The dynamic disulfide bonds were introduced through the molecular dynamics simulation of the crystal structure of KGF-2. All of the KGF-2 mutants generated by computational protein design in this study were verified with biological experiments, which revealed that K81L, K94P, K103D/C106D, V123I, F167L, A185I, H171C/Q190C mutants could increase the thermal stability of KGF-2, without affecting its biological functions.