A Greedy Two-stage Gibbs Sampling Method for Motif Discovery in Biological Sequences

For the motif discovery problem of DNA sequences, a greedy two-stage Gibbs sampling algorithm is presented, and the related software package is called Greedy MotifSAM. Based on position weight matrix (PWM) motif model, a greedy strategy for choosing the initial parameters of PWM is employed. Two sampling methods, site sampler and motif sampler, are used. Site sampler is used to find one occurrence per sequence of the motif in the dataset. Motif sampler is used to find zero or more non-overlapping occurrences of the motif in each sequence. The algorithm is capable of discovering several different motifs with differing numbers of occurrences in a single dataset. We use the binding sites (motif) information of eukaryotic transcription factors stored in TRANSFAC database to test our methods. The prediction accuracy, scalability and reliability are compared to several other methods.