Relación estructura – actividad antioxidante de derivados de cumarinas
Fecha
2018
Autores
Torres Zulueta, Claudia Daniela
Título de la revista
ISSN de la revista
Título del volumen
Editor
Universidad Central “Marta Abreu” de Las Villas. Facultad de Química Farmacia. Departamento de Licenciatura Química
Resumen
Los agentes antioxidantes tienen la función de secuestrar radicales libres reactivos, lo que retarda la oxidación de otras moléculas y disminuye los daños celulares. El método más aplicado para medir la capacidad antioxidante, es el basado en la captación del radical DPPH•. En el presente trabajo se obtuvieron modelos de regresión lineales y no lineales, que correlacionaron la estructura molecular con la capacidad secuestradora del radical DPPH• en un conjunto de 115 derivados de cumarinas. Las estructuras fueron codificadas usando descriptores bidimensionales y tridimensionales del software DRAGON; se seleccionaron los descriptores de mayor variabilidad utilizando el criterio de la entropía de Shannon. La división de las series de entrenamiento y predicción se realizó mediante el análisis de conglomerados con la finalidad de construir y validar los modelos de predicción de la actividad antiradicálica. Los modelos lineales fueron obtenidos utilizando el software MobyDigs mediante la aplicación de un algoritmo genético y los modelos no lineales fueron generados por el software WEKA utilizando máquinas de vectores soporte. El modelo lineal combinado (R2 = 81,04 y Q2 = 77,93) cumple con los parámetros establecidos para la validación interna y externa, por ello se utilizó para predecir la capacidad secuestradora a 7 patrones de cumarinas. La aplicación de técnicas no lineales mostró mejores resultados de la capacidad predictiva del modelo (R2 = 88,51 y s = 0,207). Las variables del mejor modelo de regresión lineal revelaron que existe relación entre su definición y las propiedades antioxidantes de los derivados de cumarinas.
Antioxidant agents have the function of capturing reactive free radicals, which slow down the oxidation of other molecules and reduces cell damage. The most widely applied method of measuring antioxidant capacity is based on the uptake of the DPPH• radical. In this work, linear and non-linear regression models were obtained, which correlated the molecular structure with the sequestering capacity of the radical DPPH• in a set of 115 coumarin derivatives. The structures were coded using two-dimensional and three-dimensional descriptors from the DRAGON software, the most variable descriptors were selected using the Shannon entropy criterion. The division of the training and prediction series was carried out by means of cluster analysis in order to construct and validate the prediction models of the anti-radical activity. The linear models were obtained using MobyDigs software by applying a genetic algorithm and the non-linear models were generated by WEKA software using support vector machines. The combined linear model (R2 = 81.04 and Q2 = 77.93) meets the parameters established for internal and external validation and was therefore used to predict the sequestration capacity of 7 coumarins patterns. The application of non-linear techniques shows better results in the predictive capacity of the model (R2 = 88.51 and s = 0.207). The variables of the best linear regression model revealed that there is a relationship between their definition and antioxidant properties of coumarin derivatives.
Antioxidant agents have the function of capturing reactive free radicals, which slow down the oxidation of other molecules and reduces cell damage. The most widely applied method of measuring antioxidant capacity is based on the uptake of the DPPH• radical. In this work, linear and non-linear regression models were obtained, which correlated the molecular structure with the sequestering capacity of the radical DPPH• in a set of 115 coumarin derivatives. The structures were coded using two-dimensional and three-dimensional descriptors from the DRAGON software, the most variable descriptors were selected using the Shannon entropy criterion. The division of the training and prediction series was carried out by means of cluster analysis in order to construct and validate the prediction models of the anti-radical activity. The linear models were obtained using MobyDigs software by applying a genetic algorithm and the non-linear models were generated by WEKA software using support vector machines. The combined linear model (R2 = 81.04 and Q2 = 77.93) meets the parameters established for internal and external validation and was therefore used to predict the sequestration capacity of 7 coumarins patterns. The application of non-linear techniques shows better results in the predictive capacity of the model (R2 = 88.51 and s = 0.207). The variables of the best linear regression model revealed that there is a relationship between their definition and antioxidant properties of coumarin derivatives.
Descripción
Palabras clave
Agentes Antioxidantes, Derivados de Cumarinas, Radicales Libres (RL), Estudios QSAR, Regresión Lineal, Regresión no Lineal