The development of checkpoint immunotherapy has been a paradigm shift in the treatment of cancer, leading to dramatic improvement in treatment outcomes across a broad range of tumor types. Nevertheless, our current understanding of the tumor immune microenvironment and mediators of resistance to therapy are limited. The recent development of high-throughput single-cell RNA-Sequencing (scRNA-Seq) technology has opened up an unprecedented window into the transcriptional states of distinct tumor-infiltrating immune and stromal cells. However, even this technology has its biological limitations, with very high levels of data dropout induced by low total mRNA molecules and capture efficiency. This thesis explores the application of a transcriptional regulatory protein activity inference approach to single-cell data in order to resolve gene dropout and more deeply characterize upstream drivers of cell state within the micro-environment of several distinct tumor types. To this end, algorithms for inference of protein activity, drug sensitivity, and cell-cell interaction have been adapted to scRNA-Seq data, along with an approach for querying enrichment of single-cell-derived population marker gene sets patient-by-patient in larger bulk-RNA-Seq cohorts. By applying these tools systematically, we have identified distinct cellular sub-populations associated with clinical outcome in different tumor types, including a novel population of C1Q+/TREM2+/APOE+ macrophages associated with post-surgical tumor recurrence in clear cell renal carcinoma, a sub-population of fibroblasts associated with improved response to immunotherapy in head and neck squamous cell carcinoma, tumor cell subpopulations with distinct inferred drug sensitivities in cholangiocarcinoma and prostate cancer, as well as tumor-specific regulatory T-cells (Tregs), active as a mechanism of immunotherapy resistance across a range of tumor types. In ongoing clinical trials from both primary and metastatic prostate cancer as well as clear cell renal carcinoma, we are able to assess which of these populations are enriched in non-responders to checkpoint immunotherapy. The proteomic master regulators of each of these single-cell types have direct utility as potential biomarkers for treatment response, but they may also be therapeutically modulated as novel targets for combination immunotherapy, potentially improving treatment response rates and treatment outcomes in future clinical trials. Finally, this thesis also presents a discovery-to-validation platform to accelerate micro-environment-directed drug repurposing in the context of immunotherapy resistance and rapid CRISPRko validation of novel therapeutic targets. This platform has been developed specifically to validate newly identified master regulators of tumor-specific immunosuppressive regulatory T-cells (Tregs), resulting in discovery of low-dose gemcitabine as a tumor-specific Treg-modulating drug synergistic with anti-PD1 checkpoint immunotherapy and TRPS1 as a proteomic master regulator with clinically significant effect on tumor Treg-infiltrating and tumor growth rate. However, the platform itself may be readily extended in future work to prioritize agents against immunosuppressive macrophage and fibroblast populations for clinical development and trials. As we have discovered, different cancers have different populations of cells driving therapy response and resistance. Taken together, the analytical and validation tools presented in this thesis represent an opportunity to tailor future immuno-therapies at the single-cell level to particular tumor types and to individual patients.