Classifying transcription factor targets and discovering relevant biological features

Area of Science:

• Genomics and Bioinformatics
• Molecular Biology
• Systems Biology

Background:

• Identifying transcription factor (TF) and gene interactions is crucial in post-genomic research.
• Previous work developed a supervised-learning approach for TF target identification in yeast, predicting targets for 104 TFs.
• This study incorporates new sequence conservation measures and expands predictions to 59 additional TFs.

Purpose of the Study:

• To improve TF target identification by integrating new data and methods.
• To expand TF target predictions to a larger set of yeast TFs.
• To uncover biological features contributing to gene regulation and analyze network properties.

Main Methods:

• Utilized a supervised-learning approach combining 8 genomic datasets (sequence conservation, overrepresentation, gene expression, DNA structure).
• Incorporated a new sequence conservation measure and expanded predictions to 59 new transcription factors.
• Employed recursive-feature-elimination for feature ranking to identify key regulatory features.

Main Results:

• Significantly amplified information on yeast regulators, with total target predictions exceeding known targets by over 2 for 11 TFs.
• Predicted TF targets align with known biology; Swi6 analysis suggests roles in DNA damage response and lipid metabolism.
• Identified transcriptional network hubs and genes regulated by multiple factors, highlighting roles in cell-cycle, growth, metabolism, and energy generation.

Conclusions:

• Postprocessing of regulatory classifier results yields high-quality predictions.
• Feature ranking strategies provide valuable insights into TF regulatory functions.
• Predictions and the full transcriptional network are accessible via a web server for further analysis.