Inferring the Transcriptional Modules Using Penalized Matrix Decomposition

摘要

This paper proposes to use the penalized matrix decomposition (PMD) to discover the transcriptional modules from microarray data. With the sparsity constraint on the decomposition factors, metagenes can be extracted from the gene expression data and they can well capture the intrinsic patterns of genes with the similar functions. Meanwhile, the PMD factors of each gene are good indicators of the cluster it belongs to. Compared with traditional methods, our method can cluster genes of the similar functions but without similar expression profiles. It can also assign a gene into different modules.