In silico development of anesthetics based on barbiturate and thiobarbiturate inhibition of GABAA. (October 2020)

• Record Type: Journal Article
• Authors: Stošić, Biljana; Janković, Radmilo; Stošić, Marija; Marković, Danica; Stanković, Danijela; Sokolović, Dušan; Veselinović, Aleksandar M.
• Is Part Of: Computational biology and chemistry. Volume 88(2020)
• Publication Date: 2020-10
• Subjects: GABAA inhibitors -- Anesthesia -- QSAR -- Molecular modeling -- Drug design; Chemistry -- Data processing -- Periodicals; Biology -- Data processing -- Periodicals; Biochemistry -- Data processing; Biology -- Data processing; Molecular biology -- Data processing; Periodicals; Electronic journals; 542.85
• DOI: 10.1016/j.compbiolchem.2020.107318
• Languages: English
• ISSNs: 1476-9271

Abstract

Graphical abstract: Highlights: QSAR models for GABA A inhibitory action were developed. Monte Carlo method with SMILES notation and molecular graph descriptors was used. Different methods were applied for the determination of the robustness of the model. Molecular fragments with influence on inhibitory action were determined. Presented study can be useful in the search for novel anesthetics. Abstract: The inhibition of GABAA can be used in general anesthesia. Although, barbiturates and thiobarbiturates are used in anesthesia, the mechanism of their action hasn't been established. QSAR modeling is a wieldy used technique in these cases and this study presents the QSAR modeling for a group of barbiturates and thiobarbiturates with determined anesthetic activity. Developed QSAR models were based on conformation independent and 2D descriptors as well as field contribution. As descriptors used for developing conformation independent QSAR models, (SMILES) notation and local invariants of the molecular graph were used. Monte Carlo optimization method was applied for building QSAR models for two defined activities. Methodology for developing QSAR models capable of dealing with the small dataset that integrates dataset curation, "exhaustive" double cross-validation and a set of optimal model selection techniques including consensus predictions was used. Two-dimensional descriptors with definite physicochemical meaning were used and modeling was done with the application of both partial least squares and multiple linear regression models with three latent variables related to simple and interpretable 2D descriptors. Different statistical methods, including novel method - the index of ideality of correlation, were used to test the quality of the developed models, especially robustness and predictability and all obtained results were good. In this study, obtained results indicate that there is a very good correlation between all developed models. Molecular fragments that account for the increase/decrease of a studied activity were defined and further used for the computer-aided design of new compounds as potential anesthetics.