The ability to determine whether a cancer is indolent or aggressive is highly desirable for clinical decision making. Current standard-of-care uses histological and pathological stages. Biologists have identified proliferative markers such as Ki67 to have prognostic significance across multiple solid tumors. Interestingly, the canonical oncogene MYC has been shown to increase the global gene expression level in tumor cells. However, the total number of mRNA molecules is not directly measurable, either in bulk or single-cell RNA sequencing data. Here, using a joint deconvolution model with matching bulk RNA and DNA sequencing data, we propose a novel metric, transcriptional activity score (TAS), to measure the ratio of global gene expression levels in tumor cells to that in surrounding non-tumor cells. Using data from The Cancer Genome Atlas and from the International Cancer Genome Consortium, we found that higher TAS was associated with more aggressive behavior, as defined by survival outcomes, pathologic correlates, and genomic features known to associate with aggressive behavior (e.g. genomic instability, whole genome doubling) in multiple cancer types. We further applied TAS to annotate somatic mutational events for their impact on global rather than local expression changes. In this talk, I will present the development of our transcriptome deconvolution model, DeMixT, and the subsequent development of TAS and our biological findings using the consortial datasets. In summary, we have developed a new summary metric using sequencing data from patient tumor samples, to compute, in vivo and using deconvolution, the relative global gene expression level of tumor cells. TAS may serve as a tractable phenotype to help elucidate the biology that underlies metastasis, prognosis and response to treatment in cancer.