DIpartite

DIpartite detects bipartite motifs in DNA sequences to identify transcription factor binding sites composed of two conserved regions separated by variable-length gaps.

Key Features:

• Dinucleotide Weight Matrix (DWM): Represents motifs using dinucleotide frequencies to capture interdependencies between neighboring bases, in contrast to single-position PWMs.
• Position Weight Matrix (PWM) support: Provides motif representation using traditional PWMs as an alternative to DWM for single-nucleotide position modeling.
• Gibbs Sampling Strategy: Employs Gibbs sampling for ab initio motif detection directly from raw DNA sequences.
• Shannon's Entropy Minimization: Applies Shannon entropy minimization to refine conserved motif predictions by reducing uncertainty.
• Flexible input specification: Supports specification of motif lengths, gap length, and selection between PWM and DWM representations.
• Performance Evaluation: Benchmarked against MEME, BioProspector, BiPad, and AMD using CRP datasets from Escherichia coli, sigma factors from Bacillus subtilis, and human promoter sequences, showing equivalent or improved detection of bipartite motifs with variable gaps.

Scientific Applications:

• Transcription factor binding site analysis: Identification of TFBSs that occur as two-block (bipartite) motifs separated by non-conserved spacers.
• Gene regulation studies: Analysis of regulatory elements to elucidate mechanisms of transcriptional control involving bipartite motifs.
• Comparative motif discovery: Detection of variable-gap bipartite motifs across bacterial (E. coli, B. subtilis) and human promoter datasets.

Methodology:

Ab initio motif discovery using Gibbs sampling with PWM or DWM representations and Shannon entropy minimization, modeling dinucleotide interdependencies and variable-length gaps.