Valeria Odegova , Anastasia Lavrinenko , Timur Rakhmanov , George Sysuev , Andrei Dmitrenko , Vladimir Vinogradov
{"title":"DESignSolvents:搜索和预测深共晶溶剂理化性质的开放平台","authors":"Valeria Odegova , Anastasia Lavrinenko , Timur Rakhmanov , George Sysuev , Andrei Dmitrenko , Vladimir Vinogradov","doi":"10.1039/d3gc04533a","DOIUrl":null,"url":null,"abstract":"<div><p>The use of organic solvents in various industries poses significant environmental risks. Deep eutectic solvents (DESs) have emerged as a promising alternative due to their environmentally friendly properties. However, finding a suitable DES for a specific application remains a challenge. Empirical selection has been the most prominent approach despite being resource-intensive and time-consuming. With recent advances in artificial intelligence (AI), the scientific community is presented with an opportunity to employ powerful machine learning methods to facilitate and speed up this process. In this study, we aimed to explore this opportunity in application to the design of DESs. We propose an approach to predict the physicochemical properties of DESs focusing on melting temperature, density, and viscosity. For that, we assembled a comprehensive database of two- and three-component DESs, characterized by a range of descriptors related to the three properties. We trained machine learning models on these data and evaluated their performance using cross-validation accuracy metrics. We found that gradient-boosted trees demonstrated superior performance compared to other models. With CatBoost, we achieved cross-validation <em>R</em><sup>2</sup> values of 0.76, 0.89, and 0.64, predicting melting temperature, density, and viscosity, respectively. Furthermore, we developed a web-resource, DESignSolvents, to provide users worldwide with the database of DES properties and the corresponding prediction models. We hope this resource will serve as a valuable tool for researchers and industry professionals to efficiently select and design DESs for various applications, promoting the spread of green chemistry.</p></div>","PeriodicalId":78,"journal":{"name":"Green Chemistry","volume":null,"pages":null},"PeriodicalIF":9.3000,"publicationDate":"2024-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"DESignSolvents: an open platform for the search and prediction of the physicochemical properties of deep eutectic solvents†\",\"authors\":\"Valeria Odegova , Anastasia Lavrinenko , Timur Rakhmanov , George Sysuev , Andrei Dmitrenko , Vladimir Vinogradov\",\"doi\":\"10.1039/d3gc04533a\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<div><p>The use of organic solvents in various industries poses significant environmental risks. Deep eutectic solvents (DESs) have emerged as a promising alternative due to their environmentally friendly properties. However, finding a suitable DES for a specific application remains a challenge. Empirical selection has been the most prominent approach despite being resource-intensive and time-consuming. With recent advances in artificial intelligence (AI), the scientific community is presented with an opportunity to employ powerful machine learning methods to facilitate and speed up this process. In this study, we aimed to explore this opportunity in application to the design of DESs. We propose an approach to predict the physicochemical properties of DESs focusing on melting temperature, density, and viscosity. For that, we assembled a comprehensive database of two- and three-component DESs, characterized by a range of descriptors related to the three properties. We trained machine learning models on these data and evaluated their performance using cross-validation accuracy metrics. We found that gradient-boosted trees demonstrated superior performance compared to other models. With CatBoost, we achieved cross-validation <em>R</em><sup>2</sup> values of 0.76, 0.89, and 0.64, predicting melting temperature, density, and viscosity, respectively. Furthermore, we developed a web-resource, DESignSolvents, to provide users worldwide with the database of DES properties and the corresponding prediction models. We hope this resource will serve as a valuable tool for researchers and industry professionals to efficiently select and design DESs for various applications, promoting the spread of green chemistry.</p></div>\",\"PeriodicalId\":78,\"journal\":{\"name\":\"Green Chemistry\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":9.3000,\"publicationDate\":\"2024-04-02\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Green Chemistry\",\"FirstCategoryId\":\"92\",\"ListUrlMain\":\"https://www.sciencedirect.com/org/science/article/pii/S1463926224002772\",\"RegionNum\":1,\"RegionCategory\":\"化学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"CHEMISTRY, MULTIDISCIPLINARY\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Green Chemistry","FirstCategoryId":"92","ListUrlMain":"https://www.sciencedirect.com/org/science/article/pii/S1463926224002772","RegionNum":1,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"CHEMISTRY, MULTIDISCIPLINARY","Score":null,"Total":0}
引用次数: 0
摘要
各行各业使用有机溶剂会带来巨大的环境风险。深共晶溶剂(DES)因其环境友好的特性而成为一种很有前途的替代品。然而,为特定应用寻找合适的 DES 仍是一项挑战。经验选择一直是最常用的方法,尽管这种方法耗费大量资源和时间。随着人工智能(AI)的最新进展,科学界有机会采用强大的机器学习方法来促进和加快这一过程。在本研究中,我们旨在探索这一应用于 DES 设计的机会。我们提出了一种预测 DES 理化特性的方法,重点是熔化温度、密度和粘度。为此,我们建立了一个包含双组分和三组分 DES 的综合数据库,其中包含一系列与这三种特性相关的描述符。我们在这些数据上训练了机器学习模型,并使用交叉验证准确度指标评估了这些模型的性能。我们发现,与其他模型相比,梯度提升树表现出更优越的性能。利用 CatBoost,我们在预测熔化温度、密度和粘度时的交叉验证 R2 值分别达到了 0.76、0.89 和 0.64。此外,我们还开发了一个网络资源 DESignSolvents,为全球用户提供 DES 性质数据库和相应的预测模型。我们希望这一资源能成为研究人员和业界专业人士的宝贵工具,帮助他们高效地选择和设计各种应用领域的 DES,促进绿色化学的推广。
DESignSolvents: an open platform for the search and prediction of the physicochemical properties of deep eutectic solvents†
The use of organic solvents in various industries poses significant environmental risks. Deep eutectic solvents (DESs) have emerged as a promising alternative due to their environmentally friendly properties. However, finding a suitable DES for a specific application remains a challenge. Empirical selection has been the most prominent approach despite being resource-intensive and time-consuming. With recent advances in artificial intelligence (AI), the scientific community is presented with an opportunity to employ powerful machine learning methods to facilitate and speed up this process. In this study, we aimed to explore this opportunity in application to the design of DESs. We propose an approach to predict the physicochemical properties of DESs focusing on melting temperature, density, and viscosity. For that, we assembled a comprehensive database of two- and three-component DESs, characterized by a range of descriptors related to the three properties. We trained machine learning models on these data and evaluated their performance using cross-validation accuracy metrics. We found that gradient-boosted trees demonstrated superior performance compared to other models. With CatBoost, we achieved cross-validation R2 values of 0.76, 0.89, and 0.64, predicting melting temperature, density, and viscosity, respectively. Furthermore, we developed a web-resource, DESignSolvents, to provide users worldwide with the database of DES properties and the corresponding prediction models. We hope this resource will serve as a valuable tool for researchers and industry professionals to efficiently select and design DESs for various applications, promoting the spread of green chemistry.
期刊介绍:
Green Chemistry is a journal that provides a unique forum for the publication of innovative research on the development of alternative green and sustainable technologies. The scope of Green Chemistry is based on the definition proposed by Anastas and Warner (Green Chemistry: Theory and Practice, P T Anastas and J C Warner, Oxford University Press, Oxford, 1998), which defines green chemistry as the utilisation of a set of principles that reduces or eliminates the use or generation of hazardous substances in the design, manufacture and application of chemical products. Green Chemistry aims to reduce the environmental impact of the chemical enterprise by developing a technology base that is inherently non-toxic to living things and the environment. The journal welcomes submissions on all aspects of research relating to this endeavor and publishes original and significant cutting-edge research that is likely to be of wide general appeal. For a work to be published, it must present a significant advance in green chemistry, including a comparison with existing methods and a demonstration of advantages over those methods.