Further results on message-passing algorithms for motif finding

A new class of message-passing algorithms for motif finding is presented. Motif finding is the problem of identifying a collection of common subsequences within a given set of DNA sequences. It can be cast as an integer linear program (ILP). Message-passing techniques are a computationally efficient alternative to the often infeasible combinatorial solutions to the ILP. We introduce a new graphical representation of the ILP formulation of the problem, and use it to develop new message-passing algorithms for motif finding. Simulation results demonstrate that the new algorithms have better performance and convergence properties than the previously proposed solutions.