Tumor drug response and resistance is driven by the tumor ecosystem, which includes an intricate combination of tumor cell properties and complex interactions among those cells organized in histological structures with surrounding immune and stromal cells. However, we lack a systematic framework of this ecosystem across tumor subtypes and patients upon which we can predict, study, and understand drug response in order to enable more precise diagnostics and better therapeutics. Tumor atlases at high spatial, cellular and genetic resolution provide an extraordinary opportunity to make these discoveries but require overcoming key methodological challenges. The Data Analysis Unit (DAU) will take advantage of this opportunity in the context of three metastatic cancers (melanoma, colon, and breast) and their resistance to immunotherapy or targeted therapy. The DAU will develop the next generation of computational methods to reconstruct these atlases from complex, massive, diverse and multidimensional spatial and cellular data, and ensure their immediate impact by formulating specific predictive models about drug effects and patient response. To do this, we will design adaptive power analyses for experimental design methods to drive the choice of samples, data modalities, and experimental parameters in a systematic way. We will develop approaches to quantify features from each data modality and across modalities. We will create an infrastructure to identify the scaffold of shared cellular, histological and clinical features across samples to build a tumor atlas, and illustrate the value of these atlases in discovering the mechanisms of drug resistance. Finally, we will design methods for querying, visualizing, and sharing atlas knowledge at scale to enable immediate access, partner with others in the Human Tumor Atlas Network (HTAN) and impact researchers and clinicians. Overall, the DAU will create the methodological framework to create the tumor atlases herein and for others developed in HTAN and the broader community.
