Efficient Design of Biological Experiments for Dose-Response Modeling in Nanomaterial Toxicology Studies

提案编号：1065931PI名称：冯阳评估纳米材料风险的最基本步骤之一是基于生物实验了解并正确描述其剂量-反应关系。由于成本、道德或其他资源或时间的限制，样本大小通常受到限制，有效利用可用资源至关重要。因此，实验设计，即实验剂量的选择和动物的分配，对剂量反应研究的成功起着重要的控制作用。智力优势：由于毒性数据的特殊性，有效的剂量反应建模设计尤其具有挑战性。这项研究的目的是开发一种适应剂量-效应曲线的非线性和毒性数据的方差异质性的实验设计方法，以指导生物实验中的剂量选择和动物分配，从而有效地产生剂量-效应关系。拟议的设计程序将建立在两阶段贝叶斯范式中，这为未来实验的设计提供了一种统计上有效的机制，以利用先前的信息。将确定最合适的剂量-反应和方差模型来描述毒性数据。Bootstrapping是一种计算密集的重抽样方法，与传统的统计推断方法不同，它将用于从后续实验设计所需的初步数据中获得重要信息。为了实现实际有用的设计，将同时考虑多个设计标准，并采用多目标亚启发式方法来搜索一组帕累托最优设计，以允许在实际实验环境中评估各种权衡。广泛的影响：本研究的结果将导致一个有效的实验设计程序，能够用更少的实验通过高质量的剂量-反应曲线来表征物质。从拟议的工作中获得的方法将大大减少毒理学研究中的实验成本和时间，缓解对动物伦理日益增长的担忧，并加快量化环境和职业暴露于纳米材料的风险、安全和健康影响的进展。该项目的实施将促进不同学科的融合以及大学和联邦实验室之间的合作伙伴关系。广泛的数据传播将直接影响对人类健康和环境的风险和安全评估。这一努力的教育目标是培训研究生和本科生，招募代表人数较少的学生(妇女、少数群体和低收入者)参加拟议的研究。此外，这项工作将为课堂案例研究提供宝贵的资源，并吸引年轻人才进入与工业工程(如计算机模拟、统计学等)相关的多学科领域。利用生物学和纳米技术。

Proposal No: 1065931PI Name: Feng YangOne of the most fundamental steps in assessing the risk of a nanomaterial is to understand and properly characterize its dose-response relationship based on biological experiments. Because of costs, ethics, or other limitations on resources or time, sample sizes are usually restricted and efficient use of available resources is critical. Thus, the design of experiments, i.e., the selection of experimental doses and the allocation of animals, plays an importantrole in the success of dose-response studies.Intellectual Merits: Efficient design for dose-response modeling is particularly challenging due to the special features of toxicity data. The objective of the proposed research is to develop an experimental design procedure, which accommodates the nonlinear nature of dose-response curves and variance heterogeneity of toxicity data, to guide the dose selection and animal allocation in biological experiments for the efficient generation of dose-response relationships. The proposed design procedure will be built in a two-stage Bayesian paradigm,which provides a statistically valid mechanism to utilize prior information for the design of future experiments. Most suitable dose-response as well as variance models will be identified to describe the toxicity data. Bootstrapping, a computationally intensive resampling method as opposed to conventional statistical inference methods, will be used to derive important information from the preliminary data required by the subsequent experimental design. To achieve practically useful designs, multiple design criteria will be considered simultaneously, and multi-objective metaheuristics will be adapted to search for a set of Pareto optimum designs, which allow for the evaluation of various trade-offs in practical experimental settings.Broader Impacts: The outcome of this research will result in an efficient experimental design procedure which is able to characterize substances by high-quality dose-response profiles using less experiments. The methods obtained from the proposed work will substantially reduce the experimental cost and time in toxicology studies, alleviate the rising concerns for animal ethics, and accelerate the progress toward quantifying the risk, safety and health effects of environmental and occupational exposure to nanomaterials. The implementationof the project will promote the fusion of different disciplines and the collaborative partnership between universities and federal labs. The broad data dissemination will directly impact the risk and safety assessment for human health and the environment. The educational objectives of this effort are to train graduate and undergraduate students, to recruit underrepresented (women, minority and low-income) students to participate in the proposed research. In addition, this work will provide a valuable resource for classroom case studies, and attract young talents into the multidisciplinary field interfacing industrial engineering (e.g., computer simulation, statistics, etc.) with biology and nanotechnology.