Therapeutic resistance is a major cause of patient mortality, and is nearly universal in solid tumors, including breast cancer. For example, trastuzumab (herceptin) is the archetype targeted therapy for the 20% of human epidermal growth factor receptor 2 (HER2)-positive (HER2+) breast cancer patients, but treatment only partially lowers the risk of recurrence in early stage disease, and is not curative in the advancedsetting. While both cancer stem cells (CSCs) and intra-tumor heterogeneity (ITH) are thought to contribute to tumor progression and resistance, mechanisms of resistance remain poorly characterized in the human system, and will only be addressed when resistant subclones are identified and successfully targeted. Although the apparent chaos that characterizes cancer genomes is daunting, tumors are governed by evolutionary principles that can be measured and exploited. However, quantitative approaches that account for clonal evolution, ITH, and CSCs are needed. To this end, we have developed an innovative experimental and computational framework that exploits the fact that somatically acquired report on the past proliferative historyof cancer cells and can be used to infer their subclonal architecture and evolutionary trajectories. By integrating genomic profiles from patient samples in a multi-scale model of tumor growth and statistical inference framework, this approach enables measurement of the dynamics of clonal expansions and patient-specific parameters. We hypothesize that a detailed characterization of tumor evolutionary dynamics and molecular changes in clinical samples during treatment will enable the unbiased identification of novel biomarkers and mechanisms of resistance. Given that HER2 is a validated therapeutic target for which several effective, but imperfect treatments exist, this is an excellent model in which to understand mechanisms of resistance. We propose an integrated molecular analysis of serial tissue specimens from HER2+ breast cancer patients treated in clinical trials with neoadjuvant single and dual agent HER2-targeted therapies to identify biomarkers of resistance (Aim 1). The genomic data will be analyzed in our computational framework to quantify CSC dynamics and temporal patterns of clonal evolution under treatment selective pressure (Aim 2). We will further characterize mechanisms of resistance, treatment-associated temporal molecular changes, and resistant subpopulations using patient- derived xenograft models and short-term primary patient cultures (Aim 3). By interrogating clonal evolution during therapy, our innovative approach will identify mechanisms of resistance and tumor dynamics that inform biomarker-driven treatment strategies. This strategy represents a new paradigm for treatment stratification with broad utility for other cancers.