基于机器学习的纳米材料生物效应研究的科学问题  

SCIENTIFIC QUESTIONS ON THE BIOLOGICAL EFFECTS OF NANOMATERIALS BASED ON MACHINE LEARNING

作　　者：胡献刚[1] 李佳薇 李佳蔷 靳红叶 于福波 HU Xiangang;LI Jiawei;LI Jiaqiang;JIN Hongye;YU Fubo(Key Laboratory of Pollution Processes and Environmental Criteria,Ministry of Education,College of Environmental Science and Engineering,Nankai University,Tianjin 300350,China;College of Life Sciences,Nankai University,Tianjin 300072,China)

机构地区：[1]南开大学环境科学与工程学院环境污染过程与基准教育部重点实验室,天津300350 [2]南开大学生命科学学院,天津300072

出　　处：《环境工程》2022年第6期171-181,共11页Environmental Engineering

基　　金：国家重点研发计划(2019YFC1804603)。

摘　　要：纳米材料(nanomaterials, NM)作为一类新型的化学品,认识、控制其引起的不良环境健康风险是推广应用的重要前提。近些年,机器学习(machine learning, ML)作为一种数据驱动的研究方法,在环境、化学和材料等领域引起了广泛关注。从NM引起的细胞毒性、个体效应、蛋白冠与生态冠预测等几方面,介绍了ML的应用,并分析了应用过程中,数据集、描述符、ML方法、模型可解释性等方面存在的问题和解决方法。文章提出,数据提取与挖掘方法、新描述符、新模型、模型解释方法的创新与发展将能够促进ML在NM生物效应领域的应用。同时,随着ML的发展,ML有望能够预测新型的NM和复杂的效应,并揭示了相关原理。以梳理讨论ML在NM生物效应预测中面临的重要科学问题为重点,有助于后续研究者理清思路,解决该领域的难题,促进纳米产业健康可持续发展。Nanomaterials(NM), as a new class of chemicals, recognizing and controlling their adverse environmental health risks are important for their applications. In recent years, machine learning(ML), as a data-driven approach, has attracted extensive attention in various fields, such as environment, chemistry, and materials. The review introduces the applications of ML from the aspects of NM-induced cytotoxicity, individual effects, protein corona and ecological corona predictions, and analyzes the problems and solutions during the application process from the aspects of data sets, descriptors, machine learning models, and model interpretability. The review also points out that the innovation and development of data extraction and mining methods, new descriptors, new models, and model interpretation methods will promote the application of ML in the biological effects of NM. With the development of ML, new types of NM and complex effects are expected to be predicted and their principles can be revealed. This review focuses on the discussion on the important scientific issues faced in the prediction of NM biological effects using ML, which will help the following researchers clarify the ideas, solve the problems in this field, and promote the healthy and sustainable development of the nano-industry.

关 键 词：纳米材料 生物效应 机器学习 大数据 人工智能 

分 类 号：TB383.1[一般工业技术—材料科学与工程] TP181[自动化与计算机技术—控制理论与控制工程]