Comparison of motif finders on benchmark and de novo discovery data sets.

Rank of the “correct” motif in the output of various programs. “Target” refers to data sets defined in [3]. “Ref.” gives the accession number in the JASPAR or miRbase database, or the target consensus sequence. X, no match in the top 4 results; nd, not determined (i.e. the tools were not run due to licensing restrictions on non-academic use). Results for Amadeus, Weeder, Trawler, YMF, AlignACE, and MEME are quoted from [3], as several are not freely available outside academia. Results for Dispom are quoted from [17]. “POWRS-FL” is POWRS without position sensitivity (“full length”). “Simple8” and “Simple6” are the whole-sequence, binomial-scoring algorithm described in the text, using 8-mers and 6-mers respectively. A result was considered correct if at least 6 contiguous bases of the result matched the literature motif (except ETS1 and YY1, which are effectively 4 bases long). The ranking from the more permissive PWM-based metric in [3] is shown in parenthesis where it disagrees.