Título: Aplicação de um Sistema Neural ao Problema de Classificação de Proteínas
Autores: Weinert, Wagner Rodrigo; Lopes, Heitor Silvério
Resumo: The prediction of protein function, based on its structure, is a hard problem in Molecular Biology. The number of known proteins has grown sharply, thanks to the several genomic sequencing efforts worldwide. This fact makes unfeasible the application of conventional laboratory techniques for function prediction of new proteins, therefore leading to the necessity of automatic classifiers for this purpose. Classification of an unknown protein means that it will be assigned to a known family, based on its structure (usually the primary structure), hence, supposing its function. This work presents the use of multi-layer perceptrons for protein classification. The main contribution is a novel way to encode data based on the hydrophobicity grade of amino acids of the proteomic chain. The classification of five different families are compared with a Hidden Markov Chain profile. Results show a good accuracy rate in the classification using the neural system and encourages future improvements.
Código DOI: 10.21528/CBRN2003-022
Artigo em PDF: 6CBRN_022.PDF
Arquivo BibTex: 6CBRN_022.bib