In this paper, we present a new genetic algorithm for protein structure prediction problem using face-centred cubic lattice and hydrophobic-polar energy model. Our algorithm uses i) an exhaustive generation approach to diversify the search; ii) a novel hydrophobic core-directed macro move to intensify the search; and iii) a random-walk strategy to recover from stagnation. On a set of standard benchmark proteins, our algorithm significantly outperforms the state-of-the-art algorithms for the same models.