Pathway activity levels. Schematic illustration of our approach to quantifying the activity level of a pathway. (A) A colormap of the expression levels for the genes in a hypothetical pathway after standardizing the expression levels to have zero mean and unit variance over samples. This represents the matrix Y described in the text. (B) The main component of the variation in the expression matrix depicted in (A). This representation is determined by the activity levels c and weights w (see Methods) associated with the first metagene in the singular value decomposition (SVD) of Y . The activity level in a sample (one column of the expression matrix) can be thought of as specifying a location in the range of expression profiles shown in (C). Positive activity levels here indicate relatively high (low) expression for genes with positive (negative) weight. For example, the expression profile (column) furthest to the left in the expression matrix is in the high positive region of the range of expression profiles. The colormaps in (A) and (B) show the samples divided into two hypothetical groups (e.g., case samples and control samples). We note, however, that the matrix Y contains expression values for all samples: the activity levels are determined by performing SVD using expression data from all samples without regard to how the samples are classified.