Development of a web-based predictive model of nanoparticle delivery to tumors by integrating physiologically-based pharmacokinetic modeling with artificial intelligence

项目摘要和摘要 许多研究表明，以纳米粒(NP)为基础的药物制剂在诊断和治疗中是有效的 在实验室动物中治疗癌症，但动物结果转化为临床成功的比例很低。这部分是因为 由于该领域的两个基本挑战，即NPs对肿瘤的低递送效率和缺乏 一个稳健的计算模型来解释不同物种之间的NP药代动力学(PK)差异，从而 使人们能够预测肿瘤的传播，并推断从动物到人类的结果。这样做的目的是 建议开发一种健壮的、有效的和可预测的基于生理的仿制药动学(PBPK) 雄性和雌性荷瘤小鼠的NPs模型。我们的假设是组织分布和肿瘤 通过使用数百个数据集进行训练，可以使用通用的PBPK模型来预测不同NP的交付 使用先进的数学方法，如基于贝叶斯的马尔可夫链蒙特卡罗(MCMC) 使用物种和性别特定的生理和性别的模拟和/或人工神经网络(ANN)方法 NP特有的物理化学参数。为实现这一目标制定了三个具体目标。目标1： 建立一种基于贝叶斯的稳健的通用PBPK模型，用于研究雄性和雌性荷瘤小鼠中的NPs。 目的2：建立基于贝叶斯的男性和女性NPs的稳健和预测的通用PBPK模型 通过整合人工智能来培育荷瘤小鼠。目的3：验证和优化贝叶斯-PBPK算法 用新的实验数据建立ANN模型，并将其转换为基于Web的界面。在目标1中，贝叶斯-MCMC 方法将被用来确保模型参数的严格优化和无偏。在目标2中，我们将测试 假设结合人工智能方法，如人工神经网络，将显著改善 贝叶斯-PBPK模型的预测精度、效率和适用范围。在目标3中，我们将进行 PK和荷瘤小鼠的组织分布实验来验证我们的模型。最近，我们发表了一篇 携带肿瘤的小鼠中NPs的简单PBPK模型和包含376个数据集的纳米肿瘤数据库。 这些研究使这一建议具有很高的可行性。该方案的新颖之处在于：(1)它是一种新的应用 贝叶斯-MCMC和神经网络方法在癌症纳米医学中的应用；(2)它提供了一种比较潜在性别的工具 NP肿瘤传递的差异；(3)该模型将具有预测性，这使得它不同于以前的 主要是“相关”分析的研究；及(4)该模型将转换为基于网络的界面，以 促进其在更广泛的受众中的应用。这个项目意义重大，因为它解决了一个至关重要的问题 癌症纳米药物给药效率低，这是过去20年取得进展的关键障碍 好几年了。这个项目有广泛的影响，因为它将极大地提高我们对关键 NP肿瘤交付的因素和任何潜在的性依赖，并将提供一个切实的工具，以改善 设计具有更高肿瘤递送效率的纳米粒以加速癌症纳米药物的临床转化 从动物到人类，还减少/消除了纳米医学研究中的动物实验。

PROJECT SUMMARY AND ABSTRACT Many studies have shown that nanoparticle (NP)-based drug formulations are effective in the diagnosis and treatment of cancer in lab animals, but the translation of animal results to clinical success is low. This is partly due to two fundamental challenges in this field, which are low delivery efficiency of NPs to the tumor and lack of a robust computational model to account for NP pharmacokinetic (PK) differences across species and thus allow one to predict tumor delivery and extrapolate the results from animals to humans. The objective of this proposal is to develop a robust, validated, and predictive generic physiologically based pharmacokinetic (PBPK) model for NPs in male and female tumor-bearing mice. Our hypothesis is that tissue distribution and tumor delivery of different NPs can be predicted with a generic PBPK model by training with hundreds of datasets with advanced mathematical methods, such as Bayesian-based Markov chain Monte Carlo (MCMC) simulations and/or artificial neural network (ANN) methods using species- and sex-specific physiological and NP-specific physicochemical parameters. Three specific aims were designed to achieve this objective. Aim 1: To develop a Bayesian-based robust generic PBPK model for NPs in male and female tumor-bearing mice. Aim 2: To develop a Bayesian-based robust and predictive generic PBPK model for NPs in male and female tumor-bearing mice by incorporating artificial intelligence. Aim 3: To validate and optimize the Bayesian-PBPK- ANN model with new experimental data and convert it to a web-based interface. In Aim 1, a Bayesian-MCMC method will be used to ensure model parameters are rigorously optimized and unbiased. In Aim 2, we will test the hypothesis that incorporation of artificial intelligence methods, such as ANN will significantly improve the prediction accuracy, efficiency, and applicable domain of the Bayesian-PBPK model. In Aim 3, we will conduct PK and tissue distribution experiments in tumor-bearing mice to validate our model. Recently, we published a simple PBPK model for NPs in tumor-bearing mice and a Nano-Tumor Database that contains 376 datasets. These studies make this proposal highly feasible. This project is novel because: (1) it is a new application of Bayesian-MCMC and ANN methods in cancer nanomedicine; (2) it provides a tool to compare potential sex differences in NP tumor delivery; (3) the model will be “predictive”, which makes it different from previous studies that were mostly “correlative” analysis; and (4) the model will be converted to a web-based interface to facilitate its application to a wider audience. This project is significant since it addresses a crucial problem of low delivery efficiency of cancer nanomedicines, which has been a critical barrier to progress over the last 20 years. This project has broad impacts because it will greatly improve our fundamental understanding of the key factors of NP tumor delivery and any potential sex-dependence, and will provide a tangible tool to improve the design of NPs with higher tumor delivery efficiency to accelerate clinical translation of cancer nanomedicines from animals to humans, and also reduce/eliminate animal experimentation in nanomedicine studies.