PMC:1570465 / 52985-54386 JSONTXT

{"target":"https://pubannotation.org/docs/sourcedb/PMC/sourceid/1570465","sourcedb":"PMC","sourceid":"1570465","source_url":"https://www.ncbi.nlm.nih.gov/pmc/1570465","text":"MEME reports motifs for all 36 transcription factor datasets, with e-values less than 1.0 for 20 of them. Gibbs Motif Sampler discovers motifs for 34 of the transcription factors, with 15 of them considered significant via their positive logMAP scores (no motifs are found for araC and flhCD). Projection reports motifs with no significance assessment for all 36 transcription factor datasets. The LP/DEE approach described in this paper has the best overall performance. Taking significance assessment into account, and considering all datasets with no significant motifs to have zero sSn and nPC values, our method produces 0.554 averaged sSn and 0.411 nPC values. Indeed, only two datasets, oxyr and ihf, have motifs that are deemed insignificant using our scheme yet have non-zero overlap with the actual motifs. Performance statistics for MEME and Gibbs Motif Sampler are considerably lower with the averaged sSn of 0.338 and nPC of 0.257 for Gibbs, and corresponding sSn and nPC values of 0.382 and 0.285 for MEME. Since Projection does not report significance values, we also note averages of raw coefficients for overlap with the known motifs while ignoring significance assessments. Our method still outperforms the others, though not as significantly, with sSn and nPC values of 0.565 and 0.414 for LP/DEE; 0.550 and 0.402 for Gibbs; 0.501 and 0.358 for MEME; 0.560 and 0.407 for Projection.","tracks":[]}