APPLYING ERROR-CORRECTED SEQUENCING TO DETECT MINIMAL RESIDUAL IN THE AAML 1031 STUDY

摘要 急性髓性白血病（AML）占所有儿童白血病死亡的一半，是导致儿童白血病死亡的主要原因。 成年期白血病相关死亡的原因。结果更糟的一个原因是无法正确地 评估治疗后的微小残留病（MRD）。与ALL不同，AML表现为多个 没有单一克隆表面标记的亚克隆群体，并且表面标记可以在 疗法目前AML MRD的金标准是多参数流式细胞术（MPFC）， 结果的频率为0.001，但30%的MPFC-MRD阴性患者仍然复发。或者，每个 AML病例携带白血病特异性突变，这些突变可能是疾病的标志物，但下一代除外 测序具有约1%的高错误率。在本提案中，我们将实现一种新颖的、经过验证的纠错 测序（ECS）策略，由Druley实验室与Illumina合作开发，以提高MRD 来自儿童肿瘤学的990例儿童初治AML病例中AML亚克隆异质性评估 AAML 1031组研究。我们假设使用高灵敏度的测序方法将改善 鉴定残留AML，提供关于儿科AML亚克隆异质性重要见解， 了解生殖系变异和基因功能对复发或难治性疾病的作用， 促进个性化医疗。为了质疑这一假设，我们提出以下目标： 1.在990例儿童初治患者中定义诊断和诱导1（EOI 1）结束时的亚克隆异质性 AML患者（n=1890）。通过使用有史以来最大的儿童AML前瞻性研究， 我们将对94个基因进行ECS，这些基因是儿科最常见的突变基因， 诊断时的成人AML和EOI 1，以确定与复发性疾病相关的突变模式，FAB 亚型或其他细胞遗传学特征。 2.将COG AAML 1031研究中所有受试者的ECS-MRD与现有EOI 1 MPFC-MRD相关联。 一个主要的问题是，是否“不同于正常”的细胞群被鉴定为残留疾病， MPFC实际上是ECS确定的同一人群。我们将通过ECS定义残留疾病， 将结果与MPFC状态（阳性/阴性）、实际MPFC百分比（<0.001）和临床 研究参与者的结局（复发风险、无病生存期和总生存期）。 3.将生殖系变异和所有亚克隆突变整合到机制组中， 在儿童AML中发生突变，并使用无偏见的机器学习与结果相关 算法初步数据告诉我们，每个患者在诊断和EOI时都会有多个亚克隆1 以及AML相关基因的种系变异，这可能对结果很重要。为此，我们 将考虑这些突变以及MPFC，临床特征和细胞遗传学，以确定概率 使用无监督机器学习算法进行风险评估，以改善结果预测。

Abstract Acute myeloid leukemia (AML) accounts for half of all pediatric leukemia deaths and is the leading cause of leukemia-related death in adulthood. One reason for worse outcomes is the inability to properly assess for minimal residual disease (MRD) following therapy. Unlike ALL, AML presents with multiple subclonal populations without a singular clonal surface marker, and surface markers can change during therapy. The current gold standard for AML MRD is multi-parameter flow cytometry (MPFC), which is predictive of outcomes to frequencies of 0.001, yet 30% of MPFC-MRD-negative patients still relapse. Alternatively, every AML case harbors leukemia-specific mutations that could be markers of disease, except that next-generation sequencing has high error rate of ~1%. In this proposal, we will implement a novel, validated error-corrected sequencing (ECS) strategy, developed by the Druley lab in collaboration with Illumina, to improve MRD assessment of AML subclonal heterogeneity in 990 pediatric de novo AML cases from the Children's Oncology Group AAML1031 study. We hypothesize that using a highly sensitive sequencing method will improve identification of residual AML, provide important insights on subclonal heterogeneity in pediatric AML, improve understanding of the role of germline variability and gene function on relapses or refractory disease and facilitate personalized medicine. To interrogate this hypothesis, we propose the following aims: 1. Define subclonal heterogeneity at diagnosis and end of Induction 1 (EOI1) in 990 pediatric de novo AML patients (n=1890). By using the largest prospective study of pediatric AML that has ever been performed, we will perform ECS on 94 genes that are the most frequently mutated genes in pediatric and adult AML at diagnosis and EOI1 to identify patterns of mutation associated with relapsed disease, FAB subtypes or other cytogenetic features. 2. Correlate ECS-MRD with existing EOI1 MPFC-MRD for all participants in the COG AAML1031 study. A major question is whether the “different from normal” cell population identified as residual disease by MPFC is actually the same population(s) identified by ECS. We will define residual disease by ECS and compare results to MPFC status (positive/negative), actual MPFC percentages (<0.001) and the clinical outcomes (relapse risk, disease-free survival and overall survival) of study participants. 3. Integrate germline variation and all subclonal mutations into mechanistic groups that are frequently mutated in pediatric AML and correlate with outcomes using unbiased machine learning algorithms. Preliminary data tells us that every patient will have multiple subclones at diagnosis and EOI1 as well as germline variants in AML-associated genes, which may be important for outcome. In this aim, we will take these mutations into account as well as MPFC, clinical features and cytogenetics for probabilistic risk assessment using unsupervised machine learning algorithms for improved outcome prognostication.