Machine learning prediction of nanoparticle in vitro toxicity: A comparative study of classifiers and ensemble-classifiers using the Copeland Index. (15th September 2019)

• Record Type: Journal Article
• Authors: Furxhi, Irini; Murphy, Finbarr; Mullins, Martin; Poland, Craig A.
• Is Part Of: Toxicology letters. Volume 312(2019)
• Page Start: 157
• Page End: 166
• Publication Date: 2019-09-15
• Subjects: RF Random Forest -- NN Neural Network -- BN Bayesian network -- SMO Sequential Minimal Optimization -- LR Linear Regression -- IBk instance based k-nearest neighbour -- DT Decision Table -- LWL Locally Weighted Learning -- GLM Generalized Linear Model -- SVM support vector machines -- kNN k-nearest neighbour -- SENS sensitivity -- SPEC specificity -- ACC balanced accuracy -- F1 F1-score -- DP Discriminant Power -- INT internal -- EXT external -- REL reliability -- BD Balanced Dataset -- ID Imbalanced Dataset -- NP Nano-Particle -- SIR support vector machine-instance based k-nearest neighbour-random forest -- DIR decision table-instance based k-nearest neighbour-random forest -- NIR neural network-instance based k-nearest neighbour-random forest -- LIR lazy-instance based k-nearest neighbour-random forest -- BIR Bayes-instance based k-nearest neighbour-random forest; Machine learning -- Voting -- Nanotoxicity -- Nanoparticles -- Copeland Index; Toxicology -- Periodicals; 363.179
• DOI: 10.1016/j.toxlet.2019.05.016
• Languages: English
• ISSNs: 0378-4274

Abstract

Graphical abstract: Highlights: Random Forest (RF) and Neural Network (NN) have the best performance compared to the other base classifiers. Ensemble classifiers show robustness, compared to basic classifiers, in predicting the toxicity of NP based on their properties and in vitro experimental conditions. RF and NN combined with another base classifier have not the best performance. Combining lower rank classifiers can help to catch the outliers. Copeland Index based on datasets, validation processes and performance metrics can be used to rank base and ensemble classifiers. RF, Bayesian Network (BN) and ensemble classifiers show high performances with missing values while NN did not. Abstract: Nano-Particles (NPs) are well established as important components across a broad range of products from cosmetics to electronics. Their utilization is increasing with their significant economic and societal potential yet to be fully realized. Inroads have been made in our understanding of the risks posed to human health and the environment by NPs but this area will require continuous research and monitoring. In recent years Machine Learning (ML) techniques have exploited large datasets and computation power to create breakthroughs in diverse fields from facial recognition to genomics. More recently, ML techniques have been applied to nanotoxicology with very encouraging results. In this study, categories of ML classifiers ( rules, trees, lazy, functions and bayes ) were compared using datasets from the Safe and Sustainable Nanotechnology (S2NANO) database to investigate their performance in predicting NPs in vitro toxicity. Physicochemical properties, toxicological and quantum-mechanical attributes and in vitro experimental conditions were used as input variables to predict the toxicity of NPs based on cell viability. Voting, an ensemble meta-classifier, was used to combine base models to optimize the classification prediction of toxicity. To facilitate inter-comparison, a Copeland Index was applied that ranks the classifiers according to their performance and suggested the optimal classifier. Neural Network (NN) and Random forest (RF) showed the best performance in the majority of the datasets used in this study. However, the combination of classifiers demonstrated an improved prediction resulting meta-classifier to have higher indices. This proposed Copeland Index can now be used by researchers to identify and clearly prioritize classifiers in order to achieve more accurate classification predictions for NP toxicity for a given dataset.