Cancer is thought to be caused by a sequence of multiple genetic and epigenetic alterations which occur in one or more of the genes controlling cell cycle progression and signaling transduction. The complexity of carcinogenic mechanisms leads to heterogeneity in molecular phenotype, pathology, and prognosis of cancers. Genome-wide mutational analysis of cancer genes in individual tumors is the most direct way to elucidate the complex process of disease progression, although such high-throughput sequencing technologies are not yet fully developed. As a surrogate marker for pathway activation analysis, expression profiling using microarrays has been successfully applied for the classification of tumor types, stages of tumor progression, or in some cases, prediction of clinical outcomes. However, the biological implication of those gene expression signatures is often unclear. Systems biological approaches leverage the signature genes as a representation of changes in signaling pathways, instead of interpreting the relevance between each gene and phenotype. This approach, which can be achieved by comparing the gene set or the expression profile with those of reference experiments in which a defined pathway is modulated, will improve our understanding of cancer classification, clinical outcome, and carcinogenesis. In this review, we will discuss recent studies on the development of expression signatures to monitor signaling pathway activities and how these signatures can be used to improve the identification of responders to anticancer drugs.