An efficient method to transcription factor binding sites imputation via simultaneous completion of multiple matrices...

Area of Science:

• Genomics and Molecular Biology
• Bioinformatics and Computational Biology

Background:

• Transcription factors (TFs) are crucial DNA-binding proteins regulating gene expression.
• Identifying TF binding sites is essential for understanding cellular processes and diseases.
• Experimental methods like ChIP-seq map TF binding sites genome-wide but face limitations in scope and availability.

Purpose of the Study:

• To develop a computational approach for predicting TF binding profiles.
• To overcome limitations in experimental TF mapping for integrated gene regulation analysis.
• To leverage existing ChIP-seq data for predicting missing TF binding signals.

Main Methods:

• Proposed TFBSImpute, a novel computational method for TF binding site prediction.
• TFBSImpute utilizes a 3-mode tensor fusion approach for simultaneous imputation of multiple TF binding matrices.
• Ensures positional consistency in imputed TF binding signals.

Main Results:

• Predicted TF binding signals showed significant similarity to experimental ChIP-seq data.
• TFBSImpute outperformed baseline imputation approaches in performance assessments.
• Motif analysis confirmed TFBSImpute's superior ability to capture enriched binding motifs compared to baselines.

Conclusions:

• TFBSImpute offers a valuable computational complement to experimental TF binding mapping.
• The method enhances the study of gene regulation mechanisms and disease.
• Imputed TF binding profiles provide insights beyond what is achievable through experimental data alone.