Principal component analysis (PCA) is widely used in data processing and dimensionality reduction. However, PCA su®ers from the fact that each principal component is a linear combi-nation of all the original variables, thus it is often di±cult to interpret the results. We introduce a new method called sparse principal component analysis (SPCA) using the lasso (elastic net) to produce modi¯ed principal components with sparse loadings. We show that PCA can be formulated as a regression-type optimization problem, then sparse loadings are obtained by im-posing the lasso (elastic net) constraint on the regression coe±cients. E±cient algorithms are proposed to realize SPCA for both regular multivariate data and gene expression arrays. We also give a new formula to compute the total variance of modi¯ed principal components. As illustrations, SPCA is applied to real and simulated data, and the results are encouraging.