Figure 2

Comparison between mmHG-Finder and other motif discovery tools. We evaluated the performance of mmHG-Finder in comparison to other state-of-the-art methods: MEME, DREME and XXmotif. Almost all input examples consisted of ranked lists, except for p53 (comprising target and background sets). Since MEME, DREME, and XXmotif expect to get a target set as input, we converted the ranked lists into target sets by taking the top 100 sequences for MEME (restricted by MEME’s limitation of 60,000 characters) and the top 20% sequences for the other tools. In the synthetic examples the entire ranked lists were taken as they are sufficiently small (to reflect useful comparison with MEME, as the motif is planted in top sequences, we had provided MEME, as input, with the ranking information by adding weights to the sequences, decreasing from 1 to 0 proportionally with the ranking). We used the default parameters in all comparison to other tools (e.g. zero-or-one-occurrence per sequence in MEME) and defined the expected motif length as the range 6 to 8 where possible (specifically, DREME and XXmotif do not have an input parameter for the motif length). Data and consensus motifs for p53 were taken from [31]; for REB1, CBF1, UME6, TYE7, GCN4 from [32]; and for the RNA binding proteins from [33]. Selected results are shown.