Figure 2

Relevance of copy number variables to the top CCA features from the leukemia data set. The cross-loadings, indicating the relevance of each copy number variable to the first two features extracted using regularized dual CCA, with regularization parameters τ _x = 0.9, τ _y = 0.3 (left panel), and PCA+CCA (right panel). The upper panels show the cross-loadings with the first feature, and the lower panels show the cross-loadings with the second feature. The variables are ordered along the genome and the vertical lines mark the chromosome boundaries. The cross-loadings for a copy number variable measure the degree of linear relationship between the log₂ ratio of that variable and the respective gene expression features. Hence, a higher (in absolute value) cross-loading indicates a stronger correlation between the variable and the gene expression feature. The first feature is strongly related to copy number alterations affecting chromosomes 4, 6, 8, 10, 14, 17, 18, 21 and X, while the second feature is strongly related to copy number alterations on chromosome 1 and 19, oppositely directed.