Identification of structural fingerprints for in vivo toxicity by using Monte Carlo based QSTR modeling of nitroaromatics

Abstract

Monte Carlo based method by using either SMILES based or combination of SMILES and Graph-based descriptors is an important strategy to build the QSAR/QSTR model for prediction of different biological endpoints. In this study, Monte Carlo based QSTR approach was applied to the dataset of 90 nitroaromatic compounds related to their in vivo toxicity, represented by 50% lethal dose concentration for rats (LD₅₀). Both classification and regression-based QSTR models were developed to get an idea about different fingerprints for promoters and hinderers of nitroaromatics toxicity. The best classification model was obtained by using SMILES and graph-based (GAO) descriptor with ¹EC_K connectivity (sensitivity = 0.7143, specificity = 1.0000, accuracy = 0.8889, and MCC = 0.7774). The best regression model calculated by using SMILES and hydrogen-suppressed graph descriptors with ⁰EC_k connectivity (R² = 0.7386, Q² = 0.6315, S = 0.467, and MAE = 0.340). Finally, a consensus QSTR model was generated to predict efficiently the toxicity of new compounds. The study highlighted that the comparative QSTR models by using the Monte Carlo method can also be generated and will be a useful tool for structural fingerprint analysis in case of nitroaromatics for preliminary evaluation of its toxicity to mammals.

Keywords:

• Breast cancer
• cytotoxicity
• MCF7
• phenylindole
• QSAR
• Monte Carlo

Acknowledgements

The authors are thankful to the Department of Pharmaceutical Sciences, Dr. Harisingh Gour University, Sagar, India for providing them the facilities and encouragement.

Disclosure statement

No potential conflict of interest was reported by the authors.