Tissue-specific prediction of directly regulated genes

Area of Science:

• Genomics
• Molecular Biology
• Bioinformatics

Background:

• Direct transcription factor (TF) binding to gene promoters is key evidence of gene regulation.
• Genome-wide TF binding assays are currently impractical for all cell types and conditions.
• Histone modifications correlate with cell-specific TF binding, enabling predictive modeling.

Purpose of the Study:

• To develop a more accurate method for predicting tissue-specific transcription factor (TF)-promoter interactions.
• To integrate histone modification data with traditional position weight matrix (PWM) models for enhanced prediction accuracy.

Main Methods:

• Utilized supervised learning to train a Naïve Bayes predictor for TF-promoter binding.
• Incorporated histone modification levels and PWM scores as predictor features.
• Trained and validated the predictor using TF ChIP-seq data across 23 datasets with cross-validation.

Main Results:

• The Naïve Bayes predictor incorporating histone modification (H3K4me3) and PWM scores significantly outperformed PWM scores and conservation-based models.
• Achieved higher accuracy across all sensitivity levels, with half the false positive rate compared to other methods.
• Demonstrated robust accuracy even when tested in different cell types/species and when trained without TF ChIP-seq data.

Conclusions:

• The developed predictor offers a substantial improvement in accuracy for identifying TF-promoter binding sites.
• The model's performance is robust across different biological contexts, highlighting its generalizability.
• This approach provides a practical tool for predicting gene regulation by TFs, aiding functional genomic studies.