Single-cell congruence evaluation and selection of cancer models towards precision medicine

Abstract Cancer models, such as cell lines, patient-derived xenograft (PDX) and patient-derived organoids (PDO), are instrumental substitutes for human studies in cancer research. Whether a specific cancer model mimics tumors in human patients and how to quantify the congruence are urgent and unsettled questions. With an increasing number of cancer models available, how to authenticate and select the most representative model using molecular profile information is an emerging issue. In the literature, nascent approaches are appearing through association-based or machine-learning-based approaches. While providing partial answers, existing tools have significant shortcomings that limit the potentials of rich cancer model resources towards translational and preclinical research. To decipher congruence evaluation at the level of intratumor heterogeneity and tumor microenvironment, we propose a comprehensive framework to determine molecular credentials and selection of cancer models and to guide translational and clinical applications using single-cell RNA-seq data. The specific aims are: (1) Identify and clinically annotate prevalent subclones and tumor microenvironment cell types in breast cancer. (2) Perform subclone-based congruence and selection for cancer models. (3) Decipher temporal changes in congruence of patient-derived models.