Motif discovery programs

BayesMD [1] is a probabilistic, Bayesian model for predicting novel transcription factor binding sites. Biological information about binding sites properties, background sequence models, occurrence and positional preferences are built into the model in modular fashion. Mixture prior parameters for the motif and background are trained using information on TFBSs and organismspecific promoter sequences from available databases. The main feature of this method is the use of a positional prior that provides a priori information about the location and number of occurrences of the sought motifs. In our study, the BayesMD program was run on its default parameters (2 order mixture background) and without informative positional priors, since we decided not to any mapping of data. The motif length was provided as input (w=8 for synthetic and Nanog and w=18 for p53 and ESR1).