Drug resistance is a ubiquitous problem in the therapeutic management of breast cancer, even in the context of next-generation targeted therapies where only modest clinical improvements have been observed despite a tumors mutational load for a given target pathway or intrinsic subtype. To devise effective anti-cancer treatment strategies, new systems-based methods are needed to fully interpret factors underlying drug responses encompassing both genetic and non-genetic mechanisms. Here we developed two approaches towards designing novel combination strategies for overcoming drug resistance. First, using an unbiased chemoproteomics approach, we profiled kinome dynamics across breast cancer cells in response to various targeted therapies and identified signaling changes that correlate with drug sensitivity. This signaling map identified survival factors whose presence limits the efficacy of targeted therapies and revealed AURKA as a new co-targeting opportunity to enhance the therapeutic efficacy of PI3K-pathway inhibitors in breast cancer. Second, we used single-cell transcriptomics data and pharmacogenomic modeling as a way to inform upfront drug combinations based on systematic analysis of tumor subpopulation architectures. Using in silico and experimental approaches, our study provides an effective new framework to discover drug combinations capable of counteracting intrinsic cell variability by predicting drug responses of single cells within tumor cell subpopulations and systematically links transcriptional heterogeneity with drug actionability to optimize therapy combinations.