Figure 5

Performance of transcription-prediction models built for predicted motifs. A, Models generated by motifs predicted by MAMA, existing methods, or known cis-elements of the upregulation of transcription in rice roots under iron deficiency. The best model of each method is shown. B, Best predictive-transcription models using only the presence of motifs predicted by MAMA upstream of TSS (upstream), with the presence of motifs predicted from sequences -50 bp and +150 bp from the TSS (+ near TSS) and with motif pairs upstream and near the TSS (++ pairs). C, Comparison of putative transcription models generated by each method. Model performance was measured by the area under receiver operating characteristic curves (AUC-ROC). For each method, 10 SVM runs were conducted. D, Comparison of putative transcription models between only upstream sequences (upstream) and sequences -50 bp and +150 bp from the TSS (+ near TSS) and with motif pairs (++ pairs).