This paper proposes a novel method for analyzing similarities of protein sequences based on features generated from the hydropathy properties of amino acid sequences. At first using the hydropathy characteristics, the protein sequences are converted into fixed-dimensional feature vectors. Based on these feature vectors, we analyze nine ND5 proteins and construct a phylogenic tree as an application. The phylogeny obtained is generally consistent with the phylogenic tree constructed by MEGA4 software. The experimental results show the utility of our approach.