Skip to main content
Figure 7 | BMC Bioinformatics

Figure 7

From: Network motif-based identification of transcription factor-target gene relationships by integrating multi-source biological data

Figure 7

SVM classifiers that predict the network motif for a TF on the basis of its binding site sequence and its time course gene expression profile. The figure shows the scheme for classifying TFs into four NMs. Since one TF can be assigned to more than one NM, one SVM classifier is built for each NM assignment (four classifiers in the SVM classifier block). We illuatrate the process in: (A) training phase, where the TFs with known NMs are used to train SVM classifiers, (B) operation phase, where unknown NMs are predicted by the trained SVM classifiers based on expression profile and binding site of a TF. In the training phase (A), a data set that consists of expression profile and binding site sequences is constructed for each classifier. The data set has positives (TFs with known NMs), and negatives are TFs to which randomly chosen NMs are assigned (equal in size to the positive set). The data set is used to suit the SVM classifiers. The classifiers are evaluated through the LOOCV approach (dashed box) to estimate their prediction errors. In the operation phase (B), the expression profile and binding site of a TF with unknown NM assignments are used as inputs to the SVM classifiers trained in (A). The classifiers predict the NM(s) for the TF.

Back to article page