A method for classifying a test tissue sample into a class from among K classes, where the test tissue sample has an associated gene expression vector is provided. For each class k from among the K classes, a gene expression matrix is provided based upon a gene expression level in one or more tissue samples known to be in the class k. One or more eigenvectors of the covariance matrix Cx of the centralized gene expression matrix are calculated. A metric μk that is indicative of the extent of dissimilarity between the gene expression vector and the gene expression matrix is calculated using an algebraic expression involving one or more of the eigenvectors of the matrix Cx. The test tissue sample is then classified into a class k for which the metric μk is minimum among the k classes.