Overview of de-novo motif discovery tools.

Rows indicate the learning principle, and columns indicate if the position distribution can be learned from the data. Weeder uses a consensus-based representation of the motif, while the other tools use probabilistic models. None of the existing tools is capable of searching for differentially abundant BSs and learning the positional distribution simultaneously, and developing such a tool is the goal of this work. As this tool is capable of modeling the positional preference of TFBSs using a discriminative learning principle, we call it Dispom, a tool for discriminative de-novo position distribution and motif discovery.