Improving chemical exposome target prediction by application of Coupled Matrix/Tensor-Matrix/Tensor Completion algorithms

项目摘要 麻烦被定义为公众接触的所有风险，包括 有毒化学物质。接触这些化学品对人类健康和疾病造成巨大负担。 由于时间有限，很难对所有新化学品进行全面的安全性评价， 资金然而，随着与数千次暴露相关的大量生物数据及其 分子靶点，我们假设计算方法可以发展到准确预测 新化学品的分子作用和靶点。在本建议中，我们建议实施和应用 一种新的矩阵完备化算法--耦合矩阵/张量矩阵完备化（CM/TMC） 和耦合矩阵/张量-张量完成（CM/TTC）来预测分子靶点， 大规模环境化学品暴露的靶组织。拟议的研究将是 通过以下具体目标完成：1）应用和优化CM/TMC算法， 将结果与替代方法进行比较，2）优化CM/TMC方法， 暴露靶组织预测; 3）发展CM/TTC方法用于暴露靶预测， 进行实验验证，并建立一个门户网站，用于确定目标预测。本研究 提出了第一种基于矩阵补全的曝光分子靶预测方法， 组织预测。该奖项的指导（K99）阶段的主要目标是为候选人提供 在数据科学和毒理学方面的额外培训，使他获得科学独立性， 成功实现了自己的职业目标。K99阶段将在大学进行 密歇根大学（UM），在Maureen Sartor博士、Justin Colacino博士、Kayvan Najarian博士和 马里奥·梅德韦多维奇（Mario Medvedovic），他们都是各自领域的专家。一个跨学科的顾问团队将 协助候选人的研究和职业发展。K99期工程完成后， 候选人将为成为独立调查员做好充分准备。

PROJECT SUMMARY The exposome is defined as the totality of exposures with which the public comes in contact, including toxic chemicals. Exposures to these chemicals represents a huge burden on human health and diseases. It is difficult to perform comprehensive safety assessment of all novel chemicals due to limited time and funds. However, with the vast amount of biological data related to thousands of exposures and their molecular targets, we hypothesize computational methods can be developed to accurately predict the molecular actions and targets of new chemicals. In this proposal, we propose to implement and apply a novel matrix completion algorithm named Coupled Matrix/Tensor-Matrix Completion (CM/TMC) and Coupled Matrix/Tensor-Tensor Completion (CM/TTC) to predict the molecular targets and target tissues of environmental chemical exposures at a large scale. The study proposed will be accomplished through the following specific aims: 1) Apply and optimize the CM/TMC algorithm for exposure-related datasets, comparing results to alternative methods, 2) Optimize the CM/TMC method for exposure target tissue prediction, and 3) develop CM/TTC method on exposure-target predictions, perform experimental validations, and establish a web portal for exposure-target prediction. This study poses the first matrix completion-based method on exposure molecular target predictions and target tissue predictions. The primary goal of the mentored (K99) phase of the award is to provide the candidate with additional training in data science and toxicology for him to acquire scientific independence and successfully accomplish his career objectives. The K99 phase will be conducted at the University of Michigan (UM), under the mentorship of Drs. Maureen Sartor, Justin Colacino, Kayvan Najarian, and Mario Medvedovic, who are experts in the respective fields. An interdisciplinary team of advisors will assist the candidate in his research and career development. After the completion of the K99 phase, the candidate will be well prepared to be an independent investigator.