Pro-smart : predição de estruturas terciárias de proteínas utilizando sistemas multiagente

• Autor(a) principal: Paes, Thiago Lipinski
• Data de Publicação: 2013
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da PUC_RS
• Texto Completo: http://tede2.pucrs.br/tede2/handle/tede/5236
• topic: INFORMÁTICA; SISTEMAS MULTIAGENTES; PROTEÍNAS; BIOLOGIA COMPUTACIONAL; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO

There currently are approximately 16 million of unique (non-redundant) protein sequences in the GenBank. In the PDB, we can only find about 89,000 three-dimensional (3-D) protein structures and only 1,393 different SCOP protein folds. Thus, there is a huge gap between our ability to generate protein sequences and that of solving 3-D structures of proteins with unique, novel folds. This gap has been reduced with the aid from structural bioinformatics by addressing the problem of how a protein reaches its 3-D structure starting only from its amino acid sequence. This is called the protein structure prediction (PSP) problem. Thermodynamics considerations presented by Christian Anfinsen and co-workers in 1973 have it that a protein native structure is the one that minimizes its global free energy. Hence, we can treat the PSP problem as a minimization one within an NP-complete class of computation complexity. Several techniques have been used to predict the 3-D structure of proteins. In this work we supplement these techniques by adding artificial intelligence concepts still not much exploited in bioinformatics. More specifically, we propose a framework, based on an ab initio approach, of a cooperative hierarquical multi-agent system guided by a Simulated Annealing and a Monte Carlo scheme to address the PSP problem. Our multi-agent system has as input the protein amino acid sequence. Amino acids are represented by two agents: The C-Alpha agent (in lieu of the C alpha carbon atom) and the CBeta agent (in lieu of the side chain centroid). These Amino Acid agents can interact with each other. There are two other agents: one coordinates the Amino Acid agents; the other coordinates the protein system. The multi-agent system was created using the NetLogo platform. A clustering protocol was implemented for obtaining each simulation representant model. The results were compared with published papers regarding similar methodology and the use of Multi-Agent Systems to address the Protein Structure Prediction Problem. We present partial results which are encouraging for mini proteins.