BÚSQUEDA DE ESTRUCTURAS SECUNDARIAS ÓPTIMAS Y SUBÓPTIMAS DE UNA CADENA DE ARN UTILIZANDO INTELIGENCIA ARTIFICIAL
Date
Journal Title
Journal ISSN
Volume Title
Publisher
Rev. Bol. Quim
Abstract
El ácido ribonucleico (ARN) no puede existir en forma de cadena lineal, esta molécula se repliega para alcanzar su estructura más estable mediante la formación de puentes de hidrógeno. La descripción de estos puentes es llamada, estructura secundaria del ARN. A partir de ésta, es posible deducir su estructura terciaria del ARN. En muchos casos esta estructura terciaria otorga al ARN sus propiedades, entonces resulta interesante e importante conocer la estructura secundaria de las cadenas de ARN. Encontrar éstas estructuras mediante técnicas experimentales (cristalografía de rayos X) resulta lento y largo, por ende, resulta interesante predecir éstas estructuras mediante técnicas computacionales de predicción como ser Inteligencia Artificial. En el presente trabajo se ha utilizado una red neuronal de Hopfield para encontrar diferentes estructuras secundarias estables y una red neuronal multicapa unidireccional, entrenada con ejemplos biológicos reales, para elegir la estructura secundaria más próxima a una estructura "biológica" real entre las estructuras subóptimas encontradas por la red de Hopfield. Se realizó éste trabajo (entrenamiento de la segunda red, verificación de la capacidad de predicción y validación) gracias a varias muestras de micro-ARN de drosophilas. Mediante estas redes neuronales se ha logrado predecir una estructura secundaria de ARN cercana a la "real" para un micro-ARN de Drosophila Sosophora willistoni.
Ribonucleic acid (RNA) cannot exist in the form of linear chain, this molecule folds down for achieving its structure more stable through the formation of hydrogen bridges. The description of these bridges is called, the RNA secondary structure. From this, it is possible to deduce its tertiary structure of RNA. In many cases, this tertiary structure gives to RNA its properties, then, is interesting and important to know the secondary structure of the RNA chains. Find these structures using experimental techniques (X-ray crystallography) is slow and long, therefore, it is interesting predict these structures through computational techniques of prediction such as Artificial Intelligence. In the present work has been used a Hopfield neural network to find different secondary structures stable and an unidirectional multilayer neural network trained with real biological examples, in order to choose the secondary structure more structure next to a 'biological' real structure, between suboptimal structures encountered by the Hopfield network. This work was carried out (training of the second network, verification of the ability of prediction and validation) thanks to several samples of micro-RNA of drosophilas. Through these neural networks has been predicting a RNA secondary structure close to the "real" to a micro-RNA Sosophora Drosophila willistoni.
Ribonucleic acid (RNA) cannot exist in the form of linear chain, this molecule folds down for achieving its structure more stable through the formation of hydrogen bridges. The description of these bridges is called, the RNA secondary structure. From this, it is possible to deduce its tertiary structure of RNA. In many cases, this tertiary structure gives to RNA its properties, then, is interesting and important to know the secondary structure of the RNA chains. Find these structures using experimental techniques (X-ray crystallography) is slow and long, therefore, it is interesting predict these structures through computational techniques of prediction such as Artificial Intelligence. In the present work has been used a Hopfield neural network to find different secondary structures stable and an unidirectional multilayer neural network trained with real biological examples, in order to choose the secondary structure more structure next to a 'biological' real structure, between suboptimal structures encountered by the Hopfield network. This work was carried out (training of the second network, verification of the ability of prediction and validation) thanks to several samples of micro-RNA of drosophilas. Through these neural networks has been predicting a RNA secondary structure close to the "real" to a micro-RNA Sosophora Drosophila willistoni.
Description
Vol. 29, No. 2