Cancer is the leading disease-related cause of death in children. Treatment has remained largely unchanged in decades, relying primarily on aggressive cytotoxic chemotherapy and radiation?these therapies have debilitating long-term consequences. Precision medicine has yet to make a major impact on childhood cancer because, while thousands of pediatric tumor genomes have been sequenced, most children have very few somatic mutations compared to adult cancers. This means targeted cancer drugs are not an option for most children and fewer tumor-specific neoantigens means most immunotherapies are ineffective. However, in the last 5 years, large-scale CRISPR and drug screening studies in cancer cell lines, such as the Dependency Map (DepMap), have shown that in many cancers, unmutated genes can also act as potent drug targets. These genes are known as non-oncogene dependencies. The overall goal of this project is to overcome the low mutation burden, by identifying the druggable non-oncogene dependencies of pediatric tumors and to perform the requisite in vitro and in vivo experimental work to move these therapies to the clinic. We will identify these non-oncogene dependencies by applying tools from machine learning to perform integrative analysis of large pre-clinical screening datasets (such as DepMap, CCLE, and PRISM) with patient tumor -omics data. This will allow us to nominate specific non-oncogene dependencies for pediatric tumor subtypes, defined based on, for example, whole-genome gene expression or methylation data. We will mechanistically validate the top hits using in vitro experimental assays. Additionally, almost all curative cancer treatments involve the rational combination of multiple therapies, however, existing methods to predict effective combinations perform poorly when tested on unseen data. Thus, our second aim is to apply an approach that we have developed based on targeted CRISPR knockout screening to identify synergistic drug combinations. We will validate these combinations in vivo using mouse models with patient-derived xenografts, leveraging shared resources already established at St. Jude. Finally, tumor heterogeneity is ultimately the downfall of every known cancer treatment; however, in pediatric tumors where the mutation burden is low, much of this heterogeneity is driven by cell state, rather than specific somatic mutations. We will dissect the influence of cell state on drug resistance using single-cell and spatial transcriptomics technologies applied to a drug-treated spontaneous mouse model of neuroblastoma. This will ultimately allow us to nominate new drug combinations explicitly targeting drug-resistant cell states. Overall, this research program will aim to build a pipeline at St. Jude to overcome some of the main challenges posed by the low number of somatic mutations in pediatric tumors and identify new therapeutic strategies for these patients. We have assembled a diverse world-class team of researchers with all components necessary for an eventual impact on patient care.
The treatment of pediatric cancers has remained largely unchanged in decades, relying primarily on aggressive cytotoxic chemotherapy and radiation, however, these therapies have debilitating long-term consequences. We will use new computational methodologies applied to large preclinical screening data from cancer cell lines and patient -omics datasets to investigate unmutated genes as potential drug targets in pediatric cancers (non-oncogene dependencies). We will perform the requisite downstream experimental work, including studying drug combinations and resistance mechanisms, necessary to motivate the clinical translation of our findings.
