Model-based method for transcription factor target identification with limited data.

Area of Science:

• Computational biology
• Systems biology
• Genomics

Background:

• Transcription factors (TFs) regulate gene expression by binding to DNA.
• Identifying direct TF targets is crucial for understanding gene regulatory networks.
• Existing methods often struggle with short time-series data and diverse target expression profiles.

Purpose of the Study:

• To present a novel computational method for identifying TF targets.
• To evaluate the method's performance using experimental data.
• To explore the utility of integrating spatial expression data.

Main Methods:

• A differential equation model of transcriptional regulation was fitted for each gene.
• A Gaussian process prior was used to model TF expression profiles nonparametrically.
• The method was validated using ChIP-chip and loss-of-function mutant expression data for Drosophila TFs (Twist, Mef2).

Main Results:

• The computational method successfully identified significant TF targets.
• Top-ranked genes were enriched near TF binding sites and showed differential expression in mutants.
• The model-based approach outperformed correlation-based methods for TFs with diverse expression profiles.
• Performance was comparable or superior to mutant differential expression scores.

Conclusions:

• The presented computational method is effective for identifying TF targets from wild-type gene expression time series.
• Integrating spatial expression data can further enhance target prediction accuracy.
• This approach offers a robust alternative for analyzing short time-series datasets without overfitting.