Application of Gabriella Miller Kids First Pediatric Research Data to a Predictive Model of Neuroblastoma

项目摘要 目前还没有诊断工具来准确预测儿童神经母细胞瘤疾病的预后 基于疾病的机制性质和儿童肿瘤的基因组信息。神经母细胞瘤是 交感神经系统(SNS)的实体癌性肿瘤，占所有癌症的一半。 1岁以下的婴儿。疾病发展轨迹的不确定性导致了攻击性 放射和化疗治疗往往会导致儿童长期发育障碍。 神经母细胞瘤启动的关键驱动因素的确定及其相互作用的评估 个别儿童将有助于针对化疗和限制过度治疗，可能导致增加 生活质量和婴儿存活率。 我们对这个问题的解决方案是开发一种预测性人工智能算法(Predicate NeuroB)并使用 来自儿童肿瘤的基因组输入，以测试其预测能力以预测疾病进展，识别关键 疾病驱动因素，并将结果与当前基于临床统计的算法进行比较。Predicate NeuroB基于 受体酪氨酸激酶(RTK)发育信号的网络相互作用 TrkB及其配体脑源性神经营养因子在SNS中的关键作用 发展。我们发表的模型对早期神经母细胞瘤(0-2岁婴儿)的预测 77名儿童的基因组信息比任何当前的临床预后都更准确(Kasemeier-Kulesa et Al.，2018年)。在这项研究中，我们建议加强我们的模型对更广泛类别的 应用Gabriella Miller Kids获得患者数据(年龄、疾病阶段、染色体状态、MYCN扩增) 第一个神经母细胞瘤数据库。此外，我们将在计算机上执行微扰算法以确定 能够改变神经母细胞瘤结局状态的关键驱动因素。在我们的研究结束时，通过使用 更大的患者派生数据集以及相关的临床和疾病结果信息，我们希望我们的 Predicate NeuroB模型将被证明对广泛类别的神经母细胞瘤患者和支持具有高度预测性 疾病治疗和靶向药物治疗的临床决策。

Project Summary There is currently no diagnostic tool to accurately predict pediatric neuroblastoma disease outcome that is based on the mechanistic nature of the disease and genomic information of a child’s tumor. Neuroblastoma is a solid, cancerous tumor of the sympathetic nervous system (SNS) that accounts for half of all cancers in infants younger than 1 year. Uncertainties in the trajectory of disease progression has led to aggressive radiation and chemotherapy treatments that often result in long-term developmental disabilities for children. Determination of the critical drivers of neuroblastoma initiation and assessment of their interactions for an individual child would help target chemotherapy and limit over-treatment, possibly resulting in an increased quality of life and infant survival. Our solution to this problem is to develop a predictive artificial intelligence algorithm (PredictNeuroB) and use genomic input from a child’s tumor to test its predictive strengths to predict disease progression, identify critical disease drivers and compare results to current clinical statistical-based algorithms. PredictNeuroB is based on the network interactions of receptor tyrosine kinase (RTK) developmental signals and is supported by our discovery of a critical role for trkB and its ligand brain-derived neurotrophic factor (BDNF) during SNS development. Our published model’s prediction of early stage neuroblastoma (for infants 0-2yrs old) using genomic information of 77 children is more accurate than any current clinical prognostic (Kasemeier-Kulesa et al., 2018). In this study, we propose to strengthen the predictive capability of our model for a broader class of patient data (age, stage of disease, chromosome status, MYCN amplification) by applying Gabriella Miller Kids First neuroblastoma databases. Further, we will perform in silico perturbations of the algorithm to determine critical drivers capable of altering neuroblastoma outcome states. At the conclusion of our study, by using a larger set of patient-derived data with associated clinical and disease outcome information, we expect our PredictNeuroB model will prove highly predictive for a broad class of neuroblastoma patients and support clinical decision making in disease treatment and targeted drug therapies.