In Vitro Toxicity Testing at Massive Scale in Diverse Primary Human Cells

项目概要-珊瑚基因组学是结合全基因组测序和RNA分析在细胞- 基于功能分析，以预测在代谢和毒性的化合物中遇到的人类变异， 治疗和环境暴露。暴露前后的RNA分析提供了一个高维度的 细胞功能变化的代表性，以及使用快速，具有成本效益的分析可以使毒性 在大量不同的人类细胞中进行测试。因此，为了满足毒性测试的需要， 人类种群的遗传多样性，珊瑚开发了创新的方法来降低成本， 提高高通量测序分析的速度。在成功的第一阶段SBIR（R43 HG 010445）中，珊瑚 为样品的快速浅测序建立了具有成本效益的样品多路复用协议， 重新索引工作流程，将样本汇集到单个测序运行中， 该解决方案具有新颖的功能测定和算法，以提高多基因风险评分（PRS）的性能。 在与美国食品和药物管理局（FDA）的合作中，Coral应用这些技术来预测 对乙酰氨基酚的患者特异性肝细胞反应特征（N = 200），发现个体间差异显著 毒性的变异性。重要的是，研究结果表明，患者的基因型预测了显着部分（AUC = 0.85），表明该方法对遗传多样性敏感。初步 研究结果表明，珊瑚的方法可能是一种敏感的手段，用于确定不同的毒性反应， 在不同人群中的化合物。进一步开发和测试多种化合物， 大的、多样化的样品具有提供可扩展的、高通量的平台的潜力， 毒性测试更能代表人类反应的多样性。Coral建议直接 根据NIEHS的RFA-ES-20-208进行II期SBIR，以在三种细胞模型中评价这些方法， 将至少一个模型推进到250个患者样本和100种已知毒性化合物的全面测试 数据区.目标1.表征三种人类细胞模型中RNA谱变化的个体间变异性（即， 免疫、肝细胞和胚状体）暴露于具有已知毒性特征的10种化合物。选择 Milestone：1）4，500个响应曲线; 2）个体内变异性与个体间变异性的统计学显著性 变异性（p<0.001）; 3）≥ 50%的差异表达基因与< 1 4）FAERS谱的R2 > 0.3，Tox 21数据集的AUC > 0.85。目标2.使用最具预测性的 细胞模型，表征对100种化合物的亚致死细胞毒性反应的个体间变异性， 毒性特征，以开发用于预测全身毒性的稳健模型。选择Military：1）75，000响应 特征; 2）识别≥ 5种具有高群体水平毒性变异性的化合物（第1- 50百分位数 变异>10）; 3）使用化学和转录数据预测FAERS谱的R2 > 0.4; 5）预测FAERS谱的R2 100种化合物的个体毒性特征> 0.5。

PROJECT SUMMARY—Coral Genomics is combining whole genome sequencing and RNA profiling in cell- based functional assays to predict human variation in the metabolism and toxicity of compounds encountered in therapeutic and environmental exposures. RNA profiling before and after exposure provides a high-dimensional representation of changes in cell function, and the use of rapid, cost-effective profiling could enable toxicity testing in large, diverse panels of human cells. Thus, to address the need for toxicity testing that better reflects the genetic diversity of human populations, Coral developed innovative approaches to reduce the cost and increase the speed of high-throughput sequencing assays. In a successful Phase I SBIR (R43 HG010445), Coral established protocols for cost-effective sample multiplexing for rapid shallow sequencing of samples, developed a re-indexing workflow to pool samples into a single sequencing run, and combined streamlined sequencing solutions with novel functional assays and algorithms to improve the performance of polygenic risk scores (PRS). In a collaboration with the US Food and Drug Administration (FDA), Coral applied these techniques to predict patient-specific hepatocyte response profiles to acetaminophen (N = 200) and found significant interindividual variability in toxicity. Importantly, the findings show that a patient’s genotype predicts a significant portion (AUC = 0.85) of the interindividual variation, indicating the approach is sensitive to genetic diversity. Preliminary findings indicate Coral’s approach may be a sensitive means for identifying differences in toxic responses to compounds across diverse populations. Further development and testing across multiple compounds in a large, diverse sample has the potential to provide a scalable, high-throughput platform for effective toxicity testing that is more representative of the diversity of human responses. Coral proposes a Direct to Phase II SBIR in response to NIEHS’s RFA-ES-20-208 to evaluate these methods in three cell models and advance at least one model to full-scale testing with 250 patient samples and 100 compounds with known toxicity profiles. Aim 1. Characterize the interindividual variability of RNA profile shifts in three human cell models (i.e., immune, hepatocyte, and embryoid body) exposed to ten compounds with known toxicity profiles. Select Milestones: 1) 4,500 response profiles; 2) Statistical significance of intraindividual variability vs. interindividual variability (p<0.001); 3) ≥ 50% of differentially expressed genes associated with exposure observable with < 1 million reads; 4) R2 > 0.3 for FAERS profile and AUC > 0.85 for Tox21 dataset. Aim 2. Using the most predictive cell model, characterize interindividual variability of sublethal cytotoxic responses to 100 compounds with known toxicity profiles to develop a robust model for predicting systemic toxicity. Select Milestones: 1) 75,000 response profiles; 2) Identification of ≥ 5 compounds with high population level variability in toxicity (1-50th percentile variation >10); 3) R2 > 0.4 using chemical and transcriptional data to predict FAERS profile; 5) R2 at predicting toxicity profile of an individual across 100 compounds > 0.5.