QSRR model for identification and screening of emerging pollutants based on artificial intelligence algorithms. Issue 1 (31st December 2022)

• Record Type: Journal Article
• Authors: He, Qi; Li, Hua; Jin, Binyan; Li, Wei; Shao, Bing; Zhang, Li
• Is Part Of: Environmental pollutants & bioavailability. Volume 34:Issue 1(2022)
• Page Start: 331
• Page End: 337
• Publication Date: 2022-12-31
• Subjects: QSRR -- DNN -- artificial intelligence -- relative retention time -- emerging pollutants; Pollution -- Periodicals; Bioavailability -- Periodicals; Environmental chemistry -- Periodicals; Pollution; Environmental chemistry; Bioavailability; Periodicals; Electronic journals; 577.27
• DOI: 10.1080/26395940.2022.2106311
• Languages: English
• ISSNs: 2639-5932

Abstract

ABSTRACT: It is urgent to identify and screen emerging pollutants (EPs), which have caused great harm to human health and the environment. In their detection of liquid chromatography-mass spectrometry (LC-MS), the quantitative structure–retention relationship (QSRR) model is simple and efficient to predict the retention behavior of compounds. In the present work, we collected more data with the relative retention time (RRT) of 490 compounds, and filtered the molecular descriptors with lasso regression and multiple linear regression analysis. Then ten important molecular descriptors were screened and applied the QSRR models with deep neural network (DNN), multiple linear regression (MLR), and support vector machine. The DNN model had the best accuracy which the correlation coefficient R2 reached 0.913. Finally, we determined the applicability of the DNN model through a descriptor value range to assist in the identification and screening of EPs. Graphical Abstract: uf0001