Knowledge graph embedding for profiling the interaction between transcription factors and their target genes.

Area of Science:

• Computational Biology
• Bioinformatics
• Systems Biology

Background:

• Gene regulation networks are crucial in human biology but remain complex.
• Many known transcription factor-target gene interactions lack confirmed types.
• Existing computational methods often require gene expression data, limiting their scope.

Purpose of the Study:

• To develop a novel computational model for predicting transcription factor-target gene interactions and their types.
• To leverage solely topology information for prediction, overcoming limitations of existing methods.
• To establish a benchmark dataset and evaluate the proposed model's performance.

Main Methods:

• Proposed KGE-TGI, a graph-based prediction model utilizing multi-task learning on a constructed knowledge graph.
• Formulated interaction type prediction as a multi-label classification problem on a heterogeneous graph.
• Integrated link prediction with link type classification.

Main Results:

• Achieved average AUC values of 0.9654 for link prediction and 0.9339 for link type classification in 5-fold cross-validation.
• Demonstrated that incorporating knowledge information significantly improves prediction accuracy.
• Showcased state-of-the-art performance compared to existing methodologies.

Conclusions:

• The KGE-TGI model effectively predicts gene interactions and types using topology information.
• Knowledge graph integration enhances prediction accuracy in gene regulatory network analysis.
• This approach offers a valuable tool for understanding complex biological interactions.