SOMBRERO

SOMBRERO applies a self-organizing map (SOM) neural network to discover and characterize over-represented sequence motifs using position weight matrices for analysis of biological sequences.

Key Features:

• Self-Organizing Map (SOM): Uses a SOM neural network algorithm to organize motifs based on similarity.
• Position Weight Matrices (PWMs): Represents each motif as a position weight matrix for quantitative motif characterization.
• Two-dimensional grid organization: Arranges motif PWMs on a two-dimensional grid to group related motifs spatially.
• Clustering by similarity: Clusters similar motifs together via the SOM to reveal motif families and relationships.
• Ranking by over-representation: Ranks motifs by their over-representation relative to a background model.
• Simultaneous feature characterization: Enables simultaneous characterization of all features in a dataset to reduce missed signals.
• Sensitivity to weak signals: Improves detection of weaker or subtle motifs within complex datasets.
• Comparative performance: Demonstrates improved discovery of multiple distinct motifs relative to MEME and AlignACE in reported evaluations.

Scientific Applications:

• Motif discovery in biological sequences: Identification of recurring sequence motifs within DNA or RNA datasets.
• Regulatory element analysis: Characterization of motifs associated with regulatory mechanisms and functional elements.
• Detection in complex datasets: Extraction of over-represented and weak motifs from noisy or heterogeneous sequence collections.
• Simultaneous multi-motif identification: Discovery and ranking of multiple distinct motifs present within the same dataset.

Methodology:

Organizes motifs into a two-dimensional SOM grid where each node is a position weight matrix, clusters similar motifs via the SOM algorithm, and ranks motifs by over-representation against a background model.