Consensus module eigenvectors

The module identification procedure results in modules containing CpGs with highly correlated methylation profiles. It is useful to summarize such modules using a single methylation profile per input data set. We use the module eigenvector E, defined as the left-singular vector of the standardized methylation matrix with the largest singular value[31]. Since consensus modules are defined across k independent data sets, one can form their summary profiles in each lobe. Thus, a consensus module gives rise to k eigenvectors, one in each input data set, that provide a summary “methylation value” for each sample in the data set. This allows one to relate consensus module eigenvectors to other information, for example to disease status or other traits, in each data set, and study similarities and differences between the input data sets in terms of the module-trait associations.