Redes neurais artificiais aplicadas no reconhecimento de regiões promotoras em bactérias Gram-negativas

Abstract

The promoter region is located some few base pairs before a coding region. It is responsible for initiating gene expression process, thus, it can plays a regulatory role. The study about gene expression regulation can assist mainly in the comprehension of complex metabolic network presented by several organisms and cellular answer considering the environment changes. The computational methods to gene prediction have a good accuracy, but this is not achieved in promoter prediction. This difficulty occurs because of the length of the promoter and its degenerate pattern. Those features can explain results with a great number of false positives present in the literature. The present thesis has as its main goal the neural networks applied to Gram-negative promoter prediction, recognition and characterization. Beside the classical approach with the nucleotides of the sequence, the prediction was also made by using stability values. Aiming at developing a own tool for bacterial promoter prediction, the rules extraction was carried out and the results were weighted and implemented. This tool, named BacPP, presents results comparable with the related literature. Currently, the BacPP specific accuracy for σ24, σ28, σ32, σ38, σ54 and σ70 were 86,9%; 92,8%; 91,5%; 89,3%; 97,0%; 83,6%, respectively. Furthermore, when challenged with promoter sequences belonging to other enterobacteria BacPP maintained 76% accuracy overall. Currently, there is no databases dedicated for other Gram-negative promoter than E.coli. For this reason, IntergenicDB was modeled and implemented. This database was projected to collect several pieces of information about the sequences and the organisms to which they belong and, the classification results originated from BacPP for each sequence.