Learning cell-specific networks from dynamics and geometry of single cells

Area of Science:

• Systems Biology
• Computational Biology
• Genomics

Background:

• Biological functions arise from complex molecular interaction networks.
• Inferring these networks from data is challenging, especially across diverse cell populations.
• Current methods often yield population-averaged networks, missing cell-specific interactions.

Purpose of the Study:

• To develop a method for inferring cell-specific causal gene interaction networks.
• To leverage single-cell dynamical information and cell-state manifold geometry.
• To overcome limitations of population-averaged network inference.

Main Methods:

• Introduced locaTE, an information-theoretic approach.
• Utilized single-cell dynamical data and cell-state manifold geometry.
• Developed a method agnostic to the topology of biological trajectories.

Main Results:

• Demonstrated superior performance in simulation studies.
• Successfully applied locaTE to experimental datasets (mouse development, pancreas, hematopoiesis).
• Generated additional insights compared to standard methods.

Conclusions:

• locaTE is a powerful method for inferring cell-specific networks from single-cell data.
• The approach effectively distills dynamic, cell-specific interaction information.
• Enables deeper understanding of cell heterogeneity in biological systems.