Inferência de redes gênicas com algoritmo genético e modelo de ilhas

Abstract

A massive amount of data gene expression has been produced due to the development of techniques for the extraction of molecular information, for example, the technique of RNA Seq. This development is based on the concept as the concept of central dogma of biology, in which the operation of a body is based on the expression of their genes. Knowing how is formed the structure of a regulatory gene (GRN) may contribute to a variety of applications such as understanding the operation of certain diseases, analysis of genetic diseases and to develop therapies and more effective drugs. So, computational techniques are being developed to make the inference of these networks GRNs, seeking to recover networks with high accuracy. The inference of GRNs is a challenging problem given the large amount of features (thousands of genes) and few samples (biological data). There are several methods proposed in the literature for such an inference, this paper discusses a method of feature selection. Feature selection is basically composed by a criterion function and search algorithm. The criterion function addressed in this work is based on the entropy, which aims to evaluate the possible outcomes of a given problem. And the genetic algorithm and genetic algorithm with model islands were the strategies used to perform the search of possible candidates for all gene network, and these components the target evaluation of this work. To infer and validate networks were used genetic networks Artificial (AGNs), such networks are assessable given the knowledge of the structure, and measure the effectiveness of the methods discussed. Experimental results based on the performance of the algorithms search using the model of islands obtained better results when compared to the genetic algorithm, but the computational time generated by the model of islands is higher than the runtime of the genetic algorithm.