RESUMEN
Neuroblastoma is the commonest extracranial pediatric malignancy. With few recurrent single nucleotide variations (SNVs), mutation-based precision oncology approaches have limited utility, but its frequent and heterogenous copy number variations (CNVs) could represent genomic dependencies that may be exploited for personalized therapy. Patient-derived cell culture (PDC) models can facilitate rapid testing of multiple agents to determine such individualized drug-responses. Thus, to study the relationship between individual genomic aberrations and therapeutic susceptibilities, we integrated comprehensive genomic profiling of neuroblastoma tumors with drug screening of corresponding PDCs against 418 targeted inhibitors. We quantified the strength of association between copy number and cytotoxicity, and validated significantly correlated gene-drug pairs in public data and using machine learning models. Somatic mutations were infrequent (3.1 per case), but copy number losses in 1p (31%) and 11q (38%), and gains in 17q (69%) were prevalent. Critically, in-vitro cytotoxicity significantly correlated only with CNVs, but not SNVs. Among 1278 significantly correlated gene-drug pairs, copy number of GNA13 and DNA damage response genes CBL, DNMT3A, and PPM1D were most significantly correlated with cytotoxicity; the drugs most commonly associated with these genes were PI3K/mTOR inhibitor PIK-75, and CDK inhibitors P276-00, SNS-032, AT7519, flavopiridol and dinaciclib. Predictive Markov random field models constructed from CNVs alone recapitulated the true z-score-weighted associations, with the strongest gene-drug functional interactions in subnetworks involving PI3K and JAK-STAT pathways. Together, our data defined individualized dose-dependent relationships between copy number gains of PI3K and STAT family genes particularly on 17q and susceptibility to PI3K and cell cycle agents in neuroblastoma. Integration of genomic profiling and drug screening of patient-derived models of neuroblastoma can quantitatively define copy number-dependent sensitivities to targeted inhibitors, which can guide personalized therapy for such mutationally quiet cancers.
