Mapping transcription factor binding sites by learning UV damage fingerprints

Area of Science:

• Genomics
• Molecular Biology
• Bioinformatics

Background:

• Accurate mapping of sequence-specific transcription factor binding sites (TFBS) is crucial for understanding transcriptional networks.
• Current methods like chromatin immunoprecipitation (ChIP) have limitations in identifying all TFBS.

Purpose of the Study:

• To develop and validate a novel method for identifying TFBS using machine learning and UV-induced DNA damage patterns.
• To demonstrate the utility of CPD-seq data for discovering new TFBS missed by conventional techniques.

Main Methods:

• Analysis of cyclobutane pyrimidine dimer (CPD) sequencing (CPD-seq) data using machine learning algorithms (Random Forest, neural networks).
• Identification of TFBS for specific transcription factors (Hap2/Hap3/Hap5 complex, Gcr1) in yeast and Nuclear Factor-Y in human cells.
• Validation of newly identified TFBS by association with relevant gene functions and mRNA expression levels.

Main Results:

• Identified 75 TFBS for the Hap2/Hap3/Hap5 complex in yeast, including ~25 novel sites.
• Discovered 63 Gcr1 TFBS in yeast using a neural network trained on limited known sites.
• Identified new binding sites for the Nuclear Factor-Y complex in human cells.
• Newly identified TFBS were linked to expected biological functions and altered gene expression in mutant strains.

Conclusions:

• Distinct patterns of UV damage (CPD fingerprints) at TFBS can be recognized by machine learning.
• This approach offers improved accuracy and single-nucleotide resolution for mapping TFBS.
• The method holds promise for comprehensive genome-wide identification of regulatory elements.