In this work, multivariate analysis techniques including correlation analysis, hierarchical cluster analysis (HCA), and principal component analysis (PCA) were used to estimate soybean  isoflavone genetic divergence of the 56  soybean (Glycine max) landraces from Sichuan province in China. It was found that isoflavones compositions were drastic differences among land races, and strong correlations could be observed between total isoflavone content and other isoflavones compositions. PCA showed that  three eigenvalues of  cumulative  variance proportion 91.719 %  were  selected for  evaluation of  local  soybean  varieties in  Sichuan. PC1 indicated total Isoflavone content, genistin, 6'-O-malonylgenistin, 6'-O-malonyldaidzin, daidzin were important traits for classification, while daidzein, glycitein were important in PC2. In the PC3 glycitin and 6'-O-malonylglycitin were important. Based on the results of PCA, HCA showed that all materials could be clustered into five groups, which had obvious features, so it would be helpful for parents materials selection and the isoflavone content of soybean to be improved through breeding by utilizing high daidzein or genistein or other special isoflavone composition land races