PseudoVelo: Inferring Gene Expression Derivatives Along Pseudotime as Pseudo-Velocity

Area of Science:

• Developmental Biology
• Computational Biology
• Single-cell Analysis

Background:

• Understanding cellular differentiation requires analyzing transcriptional dynamics.
• RNA velocity infers these dynamics but relies on simplified splicing kinetics.
• Existing pseudotime methods estimate transitions but bias towards mature states.

Purpose of the Study:

• Introduce PseudoVelo, a kinetics-free computational method.
• Infer gene expression derivatives (pseudo-velocity) along pseudotime.
• Provide a robust alternative for reconstructing cellular trajectories.

Main Methods:

• Utilize Generalized Additive Models to fit gene expression as a function of pseudotime.
• Employ central difference approximation to calculate gene expression derivatives.
• Generate pseudo-velocity for individual cells.

Main Results:

• PseudoVelo effectively recovers correct cellular trajectories in developmental processes.
• Demonstrates strong performance compared to existing methods like CellRank 2.
• Exhibits high resilience against data perturbations and diverse temporal priors.

Conclusions:

• PseudoVelo offers a powerful and accurate approach for inferring single-cell transcriptional dynamics.
• The method advances the analysis of cellular differentiation and transition.
• Provides a valuable tool for developmental biology research.