protein folds into a spesific native three-dimensional strecture to form its functionality. the prediction of a protein structure from its amino acid sequence is one of the most important problems in computational biology. one abstarction of the problem is so called hydrophobic-polar (HP) model, which is searching the maxiimun number of non- consecutive pairs of low free energy. A genetic algorithm search procedure was developed for us in protein folding prediction. during the generation step, a population of conformation are changed by mutation, and crossover is some parts of amino acid sequence are interchanged between conformation. the conformation of protein is represented by a sequence of moves on a cubic grid uses the HP- model. employing the GA-based technique to predict the sructures of the short proteins yield the optimal conformations with lower energy minimum thatn previously reported. further improvement to the HP-model should include other characteristics of amino acids , as the realistic structure of the protein is not necessarily the optimal structure predicted.