Дата публикации: 2020

DOI: 10.33184/bulletin-bsu-2020.4.4

С использованием программы GUSAR 2019 выполнен количественный анализ взаимосвязи между структурой и антиокислительной активностью для 113 азот-, сера- и селенсодержащих алкилфенолов в интервале значений lg k7 = 2.48÷5.30. Построено 12 статистически значимых валидных консенсус-моделей. Все эти консенсус-модели предназначены для прогноза параметра lg k7 в ряду алкилфенолов. Умеренная прогностическая способность моделей подтверждена в результате сопоставления численных значений экспериментальных и расчетных значений параметра lgk7 для структур четырех тестовых выборок, содержащих структурные аналоги моделируемых соединений.

Издатель: Башкирский государственный университет (Уфа)

Тип: Article

УДК: 544.165

Publication date: 2020

DOI: 10.33184/bulletin-bsu-2020.4.4

Using the computer program GUSAR 2019 (General Unrestricted Structure Activity Relationships), the quantitative relationship “structure - antioxidant activity (AOA)” was studied in a series of 113 nitrogen-, sulfur-, and selenium-containing alkylphenols with the general structural formulas I-IV. n= 0 ÷ 4, X= -S; -N Twelve statistically significant valid consensus models characterized by high descriptive and moderate predictive accuracy of the lgk7 parameter for the structures of training and test samples were constructed. AOA of the modeled compounds was in the range of values lgk7 = 2.48 ÷ 5.30. The descriptive and predictive ability of these models was evaluated on the structures of four training and four test samples with sliding controls with 20 times randomized exception of 20% of compounds. As statistical criteria for assessing the validity of the models M1-M9, the coefficients of determination (R20, Q2(F1), Q2(F2), , ССС) and the mean absolute error (MAE) were used. For their calculation, the XternalValidationPlus 1.2 program was used. This program showed the absence of a systematic error in the construction of all the regression dependencies M1-M12, which indicates the universality of the descriptors embedded in the GUSAR 2019 program, as well as the correctness of the regression models based on them. Thus, QSAR M1-M9 models can be successfully used for virtual screening of libraries and various databases in order to rationally search for potential antioxidants in them.

Издатель: Башкирский государственный университет (Уфа)

Тип: Article

УДК: 544.165