Each tumor is composed of diverse malignant, immune and stromal cells that interact with one another and collectively determine tumor biology and clinical phenotypes. We apply single cell RNA-seq (scRNA-seq) to diverse clinical tumor samples to comprehensively characterize the cellular diversity within tumors and explore the function of distinct tumor subpopulations. In this talk, I will describe our work on specific cancer types (e.g. glioma) and our pan-cancer integrative analysis of intra-tumor heterogeneity (ITH). We identify consistent patterns of heterogeneity across tumors, including expression meta-programs, consisting of dozens of genes that are coordinately upregulated in subpopulations of cells within many tumors. The meta-programs cover diverse cellular processes including both generic (e.g. cell cycle and stress) and lineage-specific patterns that influence metastasis, response to treatments and other tumor phenotypes.