Modelación de la capacidad de compuestos químicos de atravesar la barrera hematoencefálica
Fecha
2018
Autores
Campos Delgado, Luis Enrique
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 en Química
Resumen
En las primeras etapas del desarrollo de una nueva droga contra enfermedades del sistema nervioso central SNC, es conocido que la habilidad de los compuestos de penetrar la BHE juega un papel importante, sin embargo y aparentemente de forma paradójica, el medio interno del SNC se encuentra aislado con respecto a la circulación sanguínea por esta estructura. El objetivo fundamental de este trabajo consiste en desarrollar regresiones sobre un conjunto químico estructuralmente diverso, para encontrar relaciones lineales y no lineales que correlacionen las estructuras moleculares con sus actividades correspondientes, después de codificar numéricamente las moléculas usando índices de derivada discreta de grafos moleculares GDI. En este estudio se modeló la capacidad de atravesar le Barrera Hematoencefálica (expresada como el log BBB) de 579 entidades químicas. Las estructuras químicas fueron codificadas usando índices de derivada discreta de grafos moleculares. Estos fueron calculados utilizando el módulo DIVATI del programa TOMOCOMD-CARDD. Posteriormente, para seleccionar los descriptores con mayor variabilidad, fue utilizado el software IMMAN, que permite desarrollar análisis de variabilidad basado en el cálculo de la entropía de Shannon. Las ecuaciones de RLM fueron encontrados en el programa MobyDigs, Los modelos de regresión no lineal fueron encontrados con el software KNIME que posee implementado funciones del software WEKA, usando Máquinas de Vectores de Soporte. Se encontraron modelos de RLM (N = 560; R2 = 86.58; s = 0.565) y modelos de regresión no lineal (N= 560 R2= 80.7 s2 = 0.766). Los resultados encontrados en este estudio son superiores a los reportados en la literatura para esta misma base de datos y también para otras modelaciones que involucran la misma actividad biológica.
In early stages of the development of a new drug against diseases of the central nervous system CNS, it is known that the ability of the compounds to penetrate the BBB plays an important role, however and apparently paradoxically, the internal environment of the CNS is found isolated with respect to blood circulation by this structure. The main objective of this work is to develop regressions on a structurally diverse chemical set, to find linear and non-linear relationships that correlate the molecular structures with their corresponding activities, after coding the molecules numerically using indices of discrete derivative of molecular graphs GDI. In this study, the ability to cross the blood-brain barrier (expressed as the BBB log) of 579 chemical entities was modeled. The chemical structures were encoded using discrete derivative indices of molecular graphs. These were calculated using the DIVATI module of the TOMOCOMD-CARDD software. Subsequently, to select the descriptors with greater variability, the IMMAN software was used, which allows the development of variability analysis based on the Shannon entropy calculation. The MLR equations were found in the MobyDigs program. The non-linear regression models were found with the KNIME software that has implemented WEKA software functions, using Support Vector Machines. MLR models were found (N = 560, R2 = 86.58, s = 0.565) and non-linear regression models (N = 560 R2 = 80.7 s = 0.766). The results found in this study are superior to those reported in the literature for the same database and also for other models that involve the same biological activity.
In early stages of the development of a new drug against diseases of the central nervous system CNS, it is known that the ability of the compounds to penetrate the BBB plays an important role, however and apparently paradoxically, the internal environment of the CNS is found isolated with respect to blood circulation by this structure. The main objective of this work is to develop regressions on a structurally diverse chemical set, to find linear and non-linear relationships that correlate the molecular structures with their corresponding activities, after coding the molecules numerically using indices of discrete derivative of molecular graphs GDI. In this study, the ability to cross the blood-brain barrier (expressed as the BBB log) of 579 chemical entities was modeled. The chemical structures were encoded using discrete derivative indices of molecular graphs. These were calculated using the DIVATI module of the TOMOCOMD-CARDD software. Subsequently, to select the descriptors with greater variability, the IMMAN software was used, which allows the development of variability analysis based on the Shannon entropy calculation. The MLR equations were found in the MobyDigs program. The non-linear regression models were found with the KNIME software that has implemented WEKA software functions, using Support Vector Machines. MLR models were found (N = 560, R2 = 86.58, s = 0.565) and non-linear regression models (N = 560 R2 = 80.7 s = 0.766). The results found in this study are superior to those reported in the literature for the same database and also for other models that involve the same biological activity.
Descripción
Palabras clave
Sistema Nervioso Central (SNC), Barrera Hematoencefálica (BHE), Descriptores Moleculares, Farmácos