Predicting Relapse at the Time of Diagnosis in Acute Lymphoblastic Leukemia

项目总结 复发是白血病儿童癌症相关死亡的主要原因。尽管有了改进 在600名儿童B细胞前体急性淋巴细胞白血病(ALL)患者的总存活率中 每年会复发的人，有一半会死于他们的疾病。复发患者的高死亡率凸显了 需要改进风险预测和治疗策略，以防止白血病复发。当前 复发预测的方法受到精度不足、预测延迟和无法预测的限制 根据预测信息进行可操作的治疗调整。为了解决这些限制，我们 在诊断时对所有患者样本应用了单细胞、高参数蛋白质组学方法， 根据激活信号的前B细胞的存在，准确预测未来的复发。这 与标准护理复发预测方法相比，该方法的准确率提高了38%。我们建议 利用其独特的蛋白质组学在诊断时识别所有复发预测细胞 基因特征将导致临床风险预测模型准确、即时和 有可诉性。这种预测复发的方法将改变所有复发风险的临床范例 减少复发本身的发生率。 使用大型多机构、多模式的分子和临床注释诊断队列 患者样本，我们将应用深层蛋白质组学方法来鉴定唯一表达在 复发预测性前B细胞能够在诊断样本中直接识别。我们将决定如何 基因组突变与复发预测细胞的存在相关并检查它们的基因组 利用单细胞外显子组测序的突变负荷。最后，在我们的数据驱动的机器学习的基础上 方法，我们将构建一个比标准护理更准确的诊断复发预测因子 模型，同时告知白血病生物学和针对高危患者的针对性治疗选择。这将是 实现更精确的患者分类和治疗，减少儿童面临的 复发，并向ALL儿童的精准药物靠拢。

PROJECT SUMMARY Relapse is the major cause of cancer related mortality in children with leukemia. Despite improvements in overall survival for children with B-cell progenitor acute lymphoblastic leukemia (ALL), for the 600 patients who will relapse each year, half will die of their disease. The high mortality of patients who relapse underscores the need for improved risk prediction and treatment strategies to prevent recurrent leukemia. Current approaches to relapse prediction are limited by insufficient accuracy, delayed prediction and the inability to make actionable treatment adjustments based on prediction information. To address these limitations, we applied a single-cell, high-parameter proteomic approach to ALL patient samples at the time of diagnosis, accurately predicting future relapse based on the presence of pre-B cells with activated signaling. This approach was 38% more accurate than standard of care relapse prediction methods. We propose that identifying relapse-predictive cells in ALL at the time of diagnosis using their distinguishing proteomic and genetic features will result in a clinical risk prediction model that is accurate, immediate, and actionable. This approach to relapse prediction will change the clinical paradigm of relapse risk in ALL to reduce the incidence of relapse itself. Using large multi-institutional, multimodal cohorts of molecularly and clinically annotated diagnostic patient samples, we will apply deep proteomic approaches to identify surface proteins uniquely expressed on relapse predictive pre-B cells enabling direct identification in a diagnosis sample. We will determine how genomic mutations associate with the presence of relapse predictive cells and examine their genomic mutational burden using single-cell exome sequencing. Finally, building on our data-driven, machine learning approaches, we will construct a diagnostic relapse predictor that is more accurate than standard of care models while informing on leukemia biology and targeted therapeutic options for patients at risk. This will enable a more precise approach to patient classification and treatment, reducing the number of children facing relapse and moving closer to precision medicine for children with ALL.