Quantitative structure-antibacterial activity relationship modeling using a combination of piecewise linear regression-discriminant analysis (I): Quantum chemical, topographic, and topological descriptors

Estrada, E. and Molina, Enrique and Nodarse, Delvin and Torres, Luis A. and Gonzalez, Humberto and Uriarte, Eugenio (2008) Quantitative structure-antibacterial activity relationship modeling using a combination of piecewise linear regression-discriminant analysis (I): Quantum chemical, topographic, and topological descriptors. International Journal of Quantum Chemistry, 108 (10). pp. 1856-1871. ISSN 0020-7608 (http://dx.doi.org/10.1002/qua.21702)

Full text not available in this repository.Request a copy

Abstract

Time-dependent antibacterial activity of 2-furylethylenes using quantum chemical, topographic, and topological indices is described as inhibition of respiration in E. coli. A QSAR strategy based on the combination of the linear piecewise regression and the discriminant analysis is used to predict the biological activity values of strong and moderates antibacterial furylethylenes. The breakpoint in the values of the biological activity was detected. The biological activities of the compounds are described by two linear regression equations. A discriminant analysis is carried out to classify the compounds in one of the biological activity two groups. The results showed using different kind of descriptors were compared. In all cases the piecewise linear regression - discriminant analysis (PLR-DA) method produced significantly better QSAR models than the linear regression analysis. The QSAR models were validated using an external validation previously extracted from the original data. A prediction of reported antibacterial activity analysis was carried out showing dependence between the probability of a good classification and the experimental antibacterial activity. Statistical parameters showed the quality of quantum-chemical descriptors based models prediction in LDA having an accuracy of 0.9 and a C of 0.9. The best PLR-DA model explains more than 92% of the variance of experimental activity. Models with best prediction results were those based on quantum-chemical descriptors. An interpretation of quantum-chemical descriptors entered in models was carried out

ORCID iDs

Estrada, E. ORCID logoORCID: https://orcid.org/0000-0002-3066-7418, Molina, Enrique, Nodarse, Delvin, Torres, Luis A., Gonzalez, Humberto and Uriarte, Eugenio;
CORE (COnnecting REpositories)