Figure 5

Comparison of metamatti with MotifPrototyper and SMLR motif classification. Accuracy comparison between A) TF domain superfamily classification with metamatti and MotifPrototyper (10-fold crossvalidation) and B) TF domain family classification with metamatti and SMLR (k-fold crossvalidation). The 'major classes' in (A) refers to MotifPrototyper's reported performance for all motif families which include at least ten motif instances in TRANSFAC [37] that belong to the four superfamilies (basic, zinc, helix-turn-helix and β-scaffold), and 'full' refers to a classification of all motifs in the four superfamilies. C) The expected metamatti homeodomain motif family classification accuracy is contrasted with the accuracy achieved with M. musculus protein binding microarray derived PWMs [51] and D. melanogaster bacterial one-hybrid [52] homeodomain motif datasets. Expected accuracy refers to the out-of-bag accuracy estimate of the metamatti motif classifier.