Transcriptional programs: modelling higher order structure in transcriptional control.

Area of Science:

• Genomics
• Systems Biology
• Bioinformatics

Background:

• Transcriptional regulation is crucial in eukaryotes, but identifying transcription factor binding sites from motifs is challenging due to false positives.
• Accounting for transcription factor binding cooperativity can enhance the detection of true binding sites.

Purpose of the Study:

• To develop a non-parametric probabilistic model for detecting transcriptional programs, cooperative transcription factor groups, and co-regulated genes.
• To provide a higher-level summary of transcriptional modules and transcription factor-target gene interactions.

Main Methods:

• A non-parametric probabilistic model, analogous to a document topic model, was proposed.
• The model analyzes known transcription factor binding motifs in DNA sequences.
• It does not require prior specification of gene training sets, such as gene expression data.

Main Results:

• Applied to mouse genome data (18,445 genes), the method identified 68 transcriptional programs summarizing the activity of 149 transcription factors.
• These programs showed enrichment for known biological processes and signaling pathways.
• One program significantly overlapped with cell cycle-specific transcription factors.

Conclusions:

• The method effectively identifies higher-order structures in noisy sequence analyses, revealing potential genome-wide regulatory control mechanisms.
• It simultaneously predicts co-regulated genes and cooperating transcription factor sets.
• The discovered programs offer new avenues for studying specific transcriptional regulatory systems.