Introduction
BiPad is a stochastic greedy algorithm implemented in C++ for de novo bipartite pattern discovery, i.e. predicting bipartite motif shared by a set of nucleotide sequences BiPad web interface is written in Perl script and run bipad on the web server [1]. A biparite pattern is consisted of two adjacent one-block motifs (half-sites) separated by flexible spacing. A bipartite motif could be recognized by a transcription factor complex (3D illustration for bipartite binding). However our bipartite pattern is defined in general, (1) Gap size can be specified with any number; (2) Two half-sites could be either homogeneous (i.e. perfect/imperfect repeats) or heterogeneous (different widths/patterns); (3) Four orientations are considered: direct repeat (DR), reversed DR (RDR), inverted repeat (IR) and everted repeat (ER) or palindrome (Pal). Bipad program can search for all four orientations in an alignment.

The output is two one-block position weight matrices(PWM) for bipartite motif or one PWM matrix for one-block motif (zero-gap). Bipad takes as input the DNA sequence file in flat text (command-line) or fasta format (web interface) and search pattern parameters including motif widths and a gap range. Bipad outputs a bipartite or one-block PWM models, and coordinates and sequences of detected bipartite or one-block motifs. The web program can display motif models in graphical logos that is easy to tell the found patterns ([Bi-Logo Plotter] is also available) and a gap histogram. It can also print a motif table (click help info for more).