TGPred: efficient methods for predicting target genes of a transcription factor by integrating statistics, machine...

Area of Science:

• Computational Biology
• Bioinformatics
• Systems Biology

Background:

• Accurate inference of transcription factor (TF)-target gene (TG) interactions is crucial for understanding gene regulation.
• Existing methods for genome-wide TG prediction and pathway gene regulatory network (GRN) inference have limitations.

Purpose of the Study:

• To develop and evaluate novel statistical methods for inferring TF-TG pairs and pathway GRNs.
• To improve the speed and accuracy of computational approaches for gene regulatory analysis.

Main Methods:

• Developed four TF-TG inference methods combining MSE or Huber loss with ENET or Lasso penalties.
• Developed two pathway GRN inference methods combining Huber or MSE loss with network-based penalties.
• Utilized an accelerated proximal gradient descent (APGD) algorithm for efficient parameter optimization.

Main Results:

• ENET-based TF-TG methods outperformed Lasso-based methods on synthetic data.
• Network-based methods (Huber-Net, MSE-Net) showed superior performance for pathway GRN inference.
• Huber-ENET and MSE-ENET demonstrated high accuracy for genome-wide TF-TG predictions using transcriptomic data.

Conclusions:

• The developed TF-TG identification methods address the need for genome-wide prediction and validation of TF-TG interactions.
• The network-based methods are effective tools for predicting pathway GRNs.
• The novel methods offer enhanced speed and accuracy in computational gene regulatory network analysis.