Project Summary/Abstract An accurate tumor classification is pivotal to clinical cancer care and precision oncology. Treatment options are often informed by the pathology or diagnostic from the tumor tissue. A major challenge for patients with metastatic cancer is the limited access to tumor tissue because surgical biopsies are not routinely nor repeatedly collected throughout the course of therapy. However, tumors can undergo drastic molecular changes during metastatic progression and resistance to therapies. Circulating tumor DNA (ctDNA) released from tumor cells into the blood is a non-invasive solution for addressing challenges in tissue accessibility. Current research and clinical efforts have focused on detecting genome alterations in ctDNA, but they do not always explain treatment failure. Treatment-resistant phenotypes are defined by distinct changes in the genetic and epigenetic regulatory landscape, which collectively form the tumor regulome. Currently, it is not possible to comprehensively portray the tumor regulome in patients during the course of therapy. We propose to overcome these limitations by developing innovative computational methods and epigenetic assays that will be employed to profile the tumor regulome and survey the regulation of resistant phenotypes directly from ctDNA. Our methods will integrate the analysis of genome alterations, chromatin accessibility, transcriptional regulation, and DNA methylation from the same ctDNA sample. This cost-effective strategy provides a temporal window into the patient’s disease by monitoring the tumor epigenetic regulation and its clinical phenotype. The innovative aspects of this project include the development of deep neural networks and machine learning methods to integrate the multi-omic data extracted from a single ctDNA assay. We will employ unique systems and resources to develop our methods and advance our understanding of tumor molecular heterogeneity and treatment response. (1) From rapid autopsy studies, we will assess the contribution of DNA from multiple metastatic lesions to determine the key source of ctDNA. (2) From patient-derived xenograft (PDX) mouse models, we will establish a repository of human ctDNA from mouse plasma to support development activities and studies under PDX treatment conditions using novel therapies. This framework is generalizable to address research questions related to tumor biology and treatment response, including monitoring cancer-associated pathways and the effectiveness of targeted therapies. Successful innovations made in this project will establish a paradigm shift in cancer research and accelerate translation of new clinical applications to advance precision oncology.