Determine the protein function is important to have greater understanding of different diseases that arise from the alteration in this function. However, proteins function characterization by laboratory experiments is an expensive and long-lasting procedure. SIFTER is a bioinformatic tool used to predict protein function based on data sets obtained by experimental procedures; SIFTER uses a Hidden Markov Model (HMM), and Expectation Maximization (EM) algorithm to estimate the parameters of the Markov Model. In this paper we propose a strategy based on genetic algorithms called GAPE (Genetic Algorithm for Parameter Estimation) as an alternative to estimate the parameters of the HMM implemented in SIFTER. With the implementation of the genetic algorithm SIFTER increases the accuracy of function prediction in three of four reference data sets; in addition, computational resources (execution time and RAM memory) consume necessary for estimating the parameters of the HMM are reduced.