In this paper, a novel hybrid algorithm is represented named SA-COAMF by using cuckoo optimization algorithm, simulated annealing and expected maximization for motif finding problem. SA-COAMF is very efficient in global optimal convergence. In the algorithm, two models of motif representation, consensus and probability matrix representations, are applied to take the advantage of them. SA-COAMF is run on experimental datasets (SCPD database). The results are compared with some well-known algorithms (GADPAF, PSO+ and MEME) to show that our algorithm is efficient.