TFcomb identifies transcription factor combinations for cellular reprogramming based on single-cell multiomics data.

Area of Science:

• Computational biology
• Genomics
• Cellular reprogramming

Background:

• Cell state transitions are crucial for cell engineering and regenerative therapy.
• Identifying reprogramming transcription factors (TFs) and their combinations is essential but challenging.
• Existing computational methods often fail to identify effective TF combinations or rank individual TFs accurately.

Purpose of the Study:

• To develop a novel computational method, TFcomb, for identifying reprogramming TFs and TF combinations.
• To leverage single-cell multiomics data for improved accuracy in TF identification.
• To address limitations of current methods in identifying TF combinations and ranking individual TFs.

Main Methods:

• TFcomb models TF identification as an inverse problem, using Tikhonov regularization.
• A graph attention network augments gene regulatory networks with single-cell RNA-seq and ATAC-seq data.
• The method is validated through benchmarking on human embryonic stem cell data and diverse reprogramming cases.

Main Results:

• TFcomb demonstrates superior performance in identifying individual TFs compared to existing methods.
• The method efficiently identifies reprogramming TF combinations from extensive possibilities.
• TFcomb successfully identified key TFs in mouse hair follicle development, showcasing its applicability.

Conclusions:

• TFcomb is a powerful tool for identifying reprogramming TFs and TF combinations using single-cell data.
• The method enhances the potential for cell engineering and regenerative therapies.
• TFcomb offers significant advancements over current computational approaches for TF discovery.