Uma abordagem baseada em computação evolutiva aplicado na inferência de redes de regulação gênica

Abstract

A living organism can be seen as a network of molecules connected by biochemical reactions that has a complex system of sending and receiving signals that perform cellular control. Studies are being conducted to understand the mechanisms of control and the relationships of these reactions. They can be understood by observing the time course gene expression levels, which are estimated using methods of molecular information extraction. From these gene expression data it is possible to infer genetic regulatory networks. There are many methods to infer genetic regulatory networks in the literature, the method adopted in this work is the method of selection of features, composed of a search algorithm and a criterion function. The function criterion used is based on the average conditional entropy and the adopted is the genetic algorithm. Due to the large number of genes and the few experiments produced, the inference of genetic regulatory networks is an open challenge on bioinformatics. The discovery of relationships between the genes allows you to understand and analyze diseases, contributing to the production of more effective drugs against diseases and treatments. For inference of genetic regulatory networks have been developed two genetic algorithms, a chromosome representation for genes and the other with representation of chromosomes by networks. To validate the inferred networks used artificial gene networks. Genetic algorithm with representation of chromosomes by networks presented a greater similarity between the inferred networks and artificial gene networks, but the genetic algorithm with representation of chromosomes for genes obtained a much better computational time, thus the genetic algorithm representation of chromosome for genes has a better cost-benefit, since the difference in similarity between networks inferred by the two algorithms has been small.