Molecular simulation of the (GPx)-like antioxidant activity of ebselen derivatives through machine learning techniques

ABSTRACT

The selenoenzyme glutathione peroxidase (GPx) like activity of stable organoselenium compounds has been evaluated through the initial rate $\left({\nu }_0\right)$ of the reduction reaction of H₂O₂, Cum-OOH, and t-BuOOH. A Quantitative Structure–Activity Relationships (QSAR) analysis based on different machine learning techniques was performed by employing atom-weighed algebraic maps indexes as descriptors. The predictive capability of the obtained models was statistically validated by mean of the correlation coefficient for adjusting (R²), leave one out cross validation (Q²_LOO), and bootstrapping (Q²_boot). For the case of H₂O₂ reduction, a model was obtained with six attributes (M2) and values of R² = 0.907, Q²_LOO = 0.867, and Q²_boot = 0.852. For the cum-OOH reduction, a model was obtained with five attributes (M15) with the statistical parameters: R² = 0.925, Q²_LOO = 0.894, and Q²_boot = 0.873. For the t-BuOOH reduction, a model with four descriptors (M19) was found with the values of R² = 0.938, Q²_LOO = 0.897, Q²_boot = 0.856. The statistical parameters obtained for these three models suggest that they are robust enough with good predictive capability. Finally, screening analysis of some related compounds containing selenium was performed and two possible lead compounds were found (16 and 53), which can be used for the searching of candidates with GPx-like activity.

KEYWORDS:

• 3D-QSAR
• computational modelling
• GPx-like activity
• machine learning
• organselenium compounds

Acknowledgements

The authors are grateful to the Instituto of Biomedicina and SINDE-1831-2019 of Universidad Católica Santiago de Guayaquil (UCSG) and to the USFQ grants 2019-2021 for the financial support to this research. The High-Performance Computing system available in the USFQ was used for this project.

Disclosure statement

No potential conflict of interest was reported by the author(s).