Optimization of motif filtering for each standalone motif discovery program. The performance of each motif discovery program, applied to the Sandve et al. (2007) benchmark data set, was measured using the nucleotide Correlation Coefficient score mean (nCC, grey bar) and the nucleotide False Discovery Ratio mean (nFDR, black line). Shown is the performance of each original program (unfiltered) and after motif filtering at three probability cut-offs (p) for: (A) BioProspector, using the binomial distribution; (B) MEME using the hypergeometric distribution; and (C) Weeder using the binomial distribution. FDR and nCC error bars indicate the mean confidence intervals.