Transcriptional fidelity enhances cancer cell line selection in pediatric cancers

Multi-omic technologies have allowed for comprehensive profiling of patient-derived tumor samples and the cell lines that are intended to model them. Yet, our understanding of how cancer cell lines reflect native pediatric cancers in the age of molecular subclassification remains unclear and represents a clinical unmet need. Here we use Treehouse public data to provide an RNA-seq driven analysis of 799 cancer cell lines, focusing on how well they correlate to 1,655 pediatric tumor samples spanning 12 tumor types. For each tumor type we present a ranked list of the most representative cell lines based on correlation of their transcriptomic profiles to those of the tumor. We found that most (8/12) tumor types best correlated to a cell line of the closest matched disease type. We furthermore showed that inferred molecular subtype differences in medulloblastoma significantly impacted correlation between medulloblastoma tumor samples and cell lines. Our results are available as an interactive web application to help researchers select cancer cell lines that more faithfully recapitulate pediatric cancer.