This paper propose a least square-based sparse representation algorithm to analyze similarity comparison of protein sequences in the area of bioinformatics and molecular biology, which helps the prediction and classification of protein structure and function. The protein sequences are represented into the 1-dimensional feature vectors by their biochemical quantities. Then using the least square method to form the feature vector. Through the similarity calculation, the distance matrix can be obtained, by which, the phylogenic tree can be constructed.We apply this approach by analyzing the ND5 (NADH dehydrogenase subunit 5) protein cluster dataset. The experimental results show that the proposed model is more accurate than the Su's model,and it is closer with some known biological facts.