Identifying protein fold is an important issue in protein structure research. Based on the classification of SCOP1.65, 17 Globin-like proteins from four homology families ( < 25% sequence identity) are selected from Astral 1.65. The sequence alignment result, from structure alignment tool MUSTANG combined with manual inspection, has been used to generate a profile HMM of Globin-like fold. In a fold identify test on 68 057 sequences of Astral-1.65, the model identified 1 097 Globin-like proteins rightly, only 4 proteins of this fold are not correctly distinguished. The sensitivity and specificity of the profile HMM reach to 99.64% and 100%, respectively. Compared with Pfam and SUPERFAMILY which construct HMM based on merely sequence alignment, the model number is reduced from about 100 to 1, while keeping the sensitivity at the same level. The result shows that, for those proteins with same fold type but low sequence identity, a unified HMM could be constructed by introducing structure alignment to fold identify with high accuracy.