Identifying Combinatorial Regulatory Genes for Cell Fate Decision via Reparameterizable Subset Explanations

Area of Science:

• Developmental Biology
• Computational Biology
• Genomics

Background:

• Cell fate decisions are complex, involving intricate gene interactions.
• Disruptions in these processes can cause developmental issues and diseases.
• Existing methods struggle to model these combinatorial gene interactions effectively.

Purpose of the Study:

• Introduce MetaVelo, a framework for identifying key regulatory gene sets in cell fate transitions.
• Overcome limitations of traditional methods in capturing combinatorial gene effects.
• Provide a scalable tool for analyzing single-cell RNA sequencing data.

Main Methods:

• Model cell fate transitions as a black-box function.
• Utilize a differentiable neural ordinary differential equation (ODE) surrogate for optimization.
• Reparameterize the problem as a controllable data generation process.

Main Results:

• MetaVelo outperforms 12 baseline methods in predicting developmental trajectories.
• Successfully identifies combinatorial regulatory gene sets influencing cell fate.
• Distinguishes between independent and synergistic regulatory genes.

Conclusions:

• MetaVelo offers a superior framework for understanding cell fate dynamics.
• Provides novel insights into gene interactions critical for development and disease.
• Enables advancements in developmental biology and therapeutic applications using single-cell data.