Bioinformatics Data "Cleaning" for Immune Repertoire Sequencing

描述（由申请人提供）：本I期提案的具体目的是测试用于校正T细胞受体（TCR）库测序（REP-SEQ）中扩增偏倚的生物信息学技术的可行性，从而为该技术在临床诊断中的应用提供基础。T细胞库是人类适应性免疫的基础，并且深度T细胞库测序现在通常用于研究环境中以量化免疫应答（Robins等人，2009; Wang等人，2010; Robins等人，2012年）。临床免疫库测序具有巨大的等待市场，因为与当前的诊断方法相比，将所有可能的V（D）J组合多路复用到单个测定中显著降低了材料和劳动力成本。例如，传统的白血病微小残留病（MRD）检查需要费力的定制，每位患者的成本高达5000美元，并且周转时间为几周。我们估计，仅在MRD市场，我们的技术每年将为诊断实验室节省约1.4亿美元的成本，并将在不到标准MRD工作时间的一半的时间内产生更敏感的数据。该产品的技术创新是使用生物信息学来“清除”困扰多重库扩增的无代表性扩增（Robins等人，2012年）。首先，我们将构建TCR质粒克隆的对照文库。接下来，我们将使用对照克隆作为模板构建REP-SEQ文库，并使用下一代测序（NGS）从这些文库生成大型训练数据集。最后，我们将使用该训练集构建用于校正原始数据的线性模型，并使用第二组TCR克隆来测试线性模型的可行性。我们将要求生物信息学方法一致地清除有偏差的REP-SEQ测量，使得观察到的克隆计数与预期克隆计数之间的回归分析实现>0.95的平均R2和>0.9的平均斜率（幂=0.8，？=0.05），并且使得低至0.01%的克隆型在数百次测量中具有<10%的平均变异系数（CV（功效=0.8，<$=0.05）。此外，该技术必须足够灵敏，以可靠地检测存在的克隆型 低至百万分之一拷贝，使得在数百次测量中，受试者操作特征曲线（AUC）下的面积大于0.8（Δ =0.05）。我们在I期开发的方法将使我们能够进行大型510（k）验证研究，以获得FDA对II期临床REP-SEQ分子试剂盒的批准。最终产品的价格将低于每个样本1000美元，并将使美国各地的诊断实验室能够简化其操作，而无需将样本运送到参考实验室。 公共卫生相关性：诊断实验室经常分析T细胞以帮助表征疾病。我们正在为临床实验室的T细胞分析建立一个精简，更便宜，更全面的系统。

DESCRIPTION (provided by applicant): The Specific Aim of this Phase I proposal is to test the feasibility of a bioinformatics technology for correction of amplification bias in T cell recepor ¿ (TCR ¿) repertoire sequencing (REP-SEQ), thus providing the foundation for this technology in clinical diagnostics. The T cell repertoire is the foundation of human adaptive immunity, and deep T cell repertoire sequencing is now commonly used in a research setting to quantify immune responses (Robins et al., 2009; Wang et al., 2010; Robins et al., 2012). Clinical immune repertoire sequencing has a large awaiting market because multiplexing all possible V(D)J combinations into a single assay significantly decreases material and labor costs compared with current diagnostic methods. For example, conventional leukemia minimal residual disease (MRD) work-ups demand laborious customization, cost up to ~$5000 per patient, and have a turnaround time of several weeks. We estimate that for the MRD market alone, our technology would save ~$140 million in annual costs for diagnostics labs, and would produce more sensitive data in less than half the time of standard MRD work-ups. The technical innovation of the product is to use bioinformatics to "clean" the no representative amplification that plagues multiplexed repertoire amplification (Robins et al., 2012). First, we will build a lare control library of TCR ¿ plasmid clones. Next, we will build REP-SEQ libraries using the control clones as templates and generate a large training set of data from these libraries using next-generation sequencing (NGS). Finally, we will build a linear model for correction of raw data using this training set, and test the feasibility of the linear model using a second set of TCR ¿ clones. We will require that the bioinformatics method consistently clean biased REP-SEQ measurements such that regression analysis between observed clone counts versus expected clone counts achieves an average R2 of >0.95 and an average slope of >0.9 (power=0.8, ¿=0.05) and such that clonotypes present as low as 0.01% have an average coefficient of variation (CV) of <10% across hundreds of measurements (power=0.8, ¿ =0.05). Additionally, the technology must be sufficiently sensitive for reliable detection of clonotypes that are present as low as 1 copy in 1 million, such that the area under the receiver operator characteristic curve (AUC) is greater than 0.8 across hundreds of measurements (¿ =0.05). The methods that we develop in Phase I will enable us to perform a large 510(k) validation study for FDA approval of a molecular kit for clinical REP-SEQ in Phase II. The final product will be priced at <$1000 per sample and will enable diagnostics labs throughout the US to streamline their operations without having to ship samples to a reference lab. PUBLIC HEALTH RELEVANCE: Diagnostics laboratories often analyze T cells to help characterize disease. We are building a streamlined, cheaper, and more comprehensive system for T cell analysis in clinical laboratories.