where X is the set of sequences in the dataset, Pr(X|φ, Z) is the likelihood of the sequences X given the motif model φ and its binding sites Z, and Pr(X|p0) gives the likelihood of the sequences assuming the background model p0.