Premium
Toxicity modelling of nanomaterials by origin evaluation of their physicochemical descriptors using a combination of principal component analysis and support vector machine methods
Author(s) -
Jha Sunil Kr,
Yoon Tae Hyun
Publication year - 2020
Publication title -
expert systems
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.365
H-Index - 38
eISSN - 1468-0394
pISSN - 0266-4720
DOI - 10.1111/exsy.12492
Subject(s) - principal component analysis , support vector machine , computer science , nanomaterials , artificial intelligence , pattern recognition (psychology) , cluster analysis , classifier (uml) , machine learning , materials science , nanotechnology
In the present study, the performance of physicochemical descriptors of metal oxide nanomaterials on the basis of their origins, including descriptors related to element/ion, the metal oxide in bulk, and metal oxide in the media for toxicity modelling has been evaluated. Three published experimental nanomaterial data sets were selected for the study. The data set was divided into three subsets on the basis of the origin of descriptors; thereafter, each of them was analysed using principal component analysis for visual discrimination of toxic versus nontoxic nanomaterials in the principal component (PC) space. The metal oxide in media‐based descriptor subset results in the best visual clustering of toxicity of nanomaterials compared with the rest two subsets. It was also confirmed with the class separability measures in the PC space and classification accuracy of the support vector machine (SVM) method. PC scores of the metal oxide in media‐related descriptors results in the maximum value of class separability index ( J = 0.0049) and the maximum classification accuracy of 96.43% of SVM classifier (sensitivity of 100%). A toxicity classification model of nanomaterials has been established using PC scores of optimal descriptor subset and SVM method.