An in silico approach to cytotoxicity of pharmaceuticals and personal care products on the rainbow trout liver cell line RTL‐W1

Abstract

The authors constructed novel, robust, and validated linear Quantitative Structure-Toxicity Relationship (QSTR) models in line with Organisation of Co-operation and Development (OECD) criteria using 2 cytotoxicity data sets which were obtained from the Alamar Blue and 5-carboxyfluorescein diacetate acetoxymethyl ester (CFDA-AM) assays. The data sets comprise the cytotoxic effect of structurally diverse and widely used pharmaceuticals, synthetic musks, and industrial chemicals on the rainbow trout (Oncorhynchus mykiss) liver cell line RTL-W1. Common descriptors defined the relationship between structure and cytotoxicity for both the Alamar Blue and the CFDA-AM assays which measure the metabolic activity and membrane integrity, respectively. Only the statistical parameters of the best Alamar Blue-based model were given (n = 13; R = 0.839; the root-mean-square error of the training set [RMSE ] = 0.261; n = 5; R = 0.903; RMSE = 0.181; CCC = 0.939). The proposed QSTR model was able to predict the cytotoxicity of 101 diverse chemicals on the RTL-W1 cell line with 91% structural coverage. The authors found that in vitro-derived cytotoxicity data are promising predictors of in vivo fish toxicity and may provide an initial, rapid screening tool for acute fish toxicity assessment and reduce the need for extensive in vivo toxicity testing. Environ Toxicol Chem 2017;36:1162-1169. © 2016 SETAC.

Regression Quantitative Structure-toxicity Relationship of Pesticides on Fishes

“…3 The optimal GRNN model (σ = 0.11) yielded coefficients of determination R 2 = 0.7922 and rms = 0.6910 log units for the training set (774 samples), R 2 = 0.7278 and rms = 0.7486 log units for the validation set (166 samples), R 2 = 0.7748 and rms = 0.6903 log units for the test set (166 samples). Although the GRNN model dealt with a large dataset of pesticide toxicity logLC 50 to fishes, it is comparable to the latest similar models from the literature that have the number of samples and R 2 for the training sets being n = 13 and R 2 = 0.839 (Önlü & Saçan 2017), n = 94 and R 2 = 0.79 (Toropov et al 2017), n = 66 and R 2 = 0.80 (Khan et al 2019) n = 249 and R 2 = 0.80 (Jia et al 2020) and n = 233 and R 2 = 0.67 (Toropov et al 2020).…”

Cheminformatics and Machine Learning Approaches to Assess Aquatic Toxicity Profiles of Fullerene Derivatives

“…Computational methods have been developed over many years and play a critical role in estimating the aquatic toxicity of both new and existing compounds [ 33 , 34 , 35 , 36 , 37 , 38 , 39 ]. Quantitative structure–activity relationships (QSARs) are an effective method for predicting the aquatic toxicity of chemicals [ 40 , 41 , 42 , 43 , 44 , 45 ].…”

Norm Index–Based QSAR Model for Acute Toxicity of Pesticides Toward Rainbow Trout

“…An efficient and economical alternative approach is urgently needed to evaluate whether compounds are hazardous (Jia et al ). As a consequence, it is essential to develop in silico methods such as the quantitative structure‐activity relationships (QSARs) approach, to fill data gaps (Claeys et al ; Martínez‐López et al ; Önlü and Saçan ).…”