Motif Enrichment Analysis: a unified framework and an evaluation on ChIP data

Area of Science:

• Molecular Biology
• Bioinformatics
• Genomics

Background:

• Determining gene transcription mechanisms is crucial in molecular biology.
• Motif Enrichment Analysis (MEA) identifies DNA-binding transcription factors regulating gene sets by detecting motif enrichment in regulatory regions.
• High-throughput technologies are increasing the number of known transcription factor binding models, enhancing MEA's utility.

Purpose of the Study:

• To explore new ways to apply MEA in broader settings.
• To evaluate the efficacy of various MEA approaches.
• To develop improved MEA methods that relax input requirements.

Main Methods:

• Developed a mathematical framework for MEA using regulatory regions labeled with biological signal levels, removing the need for pre-selected gene sets.
• Implemented and evaluated several MEA methods, including user-specified threshold, data-driven threshold, and threshold-free approaches.
• Compared novel methods against existing tools (Clover, PASTAA) using yeast ChIP-chip data.

Main Results:

• A novel threshold-free linear regression method demonstrated superior performance in MEA.
• The data-driven PASTAA algorithm was the second-best performing method.
• Existing methods like Clover performed well only with optimal user-specified thresholds, while other data-driven methods performed poorly, especially without threshold limitations.

Conclusions:

• The novel, threshold-free linear regression method (AME) is effective for MEA on ChIP-chip data.
• Data-driven threshold determination can be unreliable without prior range limitations; PASTAA's limitations appear effective.
• The new MEA algorithms, AME, are publicly available for use.