Fig. 3

Evaluation of SMORE’s performance on synthetic data with known ground truth. a Accuracy of SMORE, defined as the best correlation coefficient between the embedded motif and extracted motifs for length 4 motifs with varying embedding frequency (f). This figure represents the average results from 100 algorithm runs. The highlighted curves indicate the best correlation among the n = 10 output results. Faded curves demonstrate the results when considering only n = 1, 2, or 3 of the best output motifs, instead of all 10. Examples of motifs with different levels of correlation are shown on the top. b Specificity results for length 4 motifs, measured by false positive rate and true positive rates against log p value threshold and (c) against each other. d Sensitivity of SMORE measured by successful motif recovery rate against embedding frequency and (e) enrichment log p value (see Additional file 1: Fig. S3 for length 5 results)