Utilização de máquinas de suporte vetorial para predição de estruturas terciárias de proteínas / Support vector machine for tertiary structure prediction

AUTOR(ES)
DATA DE PUBLICAÇÃO

2007

RESUMO

The three-dimensional structure of a protein is directly related to its function. Many projects of genetic sequence analysis accumulate a great number of protein sequences whose primary and secondary structures are known. However, the information on its three-dimensional structures are available only for a small fraction of these proteins. This fact evidences the necessity of creation of automatic methods for the prediction of tertiary protein structures from its primary structures. Consequently, computational tools are used for the treatment, election and analysis of these data. Currently, a new method of machine learning called Support Vector Machine (SVM) has surpassed traditional methods as Artificial Neural Networks (ANN) in the treatment of classication problems. In this master thesis we use the SVM for the automatic protein classication. The main contribution of this work was the methodology proposal for the treatment of the problem. This methodology consists in composing the support vectors with the values of the predicted secondary structures alignment of the selected proteins set. In this work, we analyze a set of data related to the 27 more popular folds described in the SCOP hierarchy. The data had been trained with diferent attributes, reaching a tax of 57% of examples classied correctly for the optimum model

ASSUNTO(S)

biologia molecular bioinformática tertiary protein structures machine learning máquina de suporte vetorial ciencia da computacao support vector machine computação

ACESSO AO ARTIGO

Documentos Relacionados