Generalizing RNA velocity to transient cell states through dynamical modeling.

Area of Science:

• Computational Biology
• Genomics
• Molecular Biology

Background:

• RNA velocity analyzes cellular differentiation using spliced and unspliced mRNA ratios in single-cell RNA sequencing (scRNA-seq).
• Limitations exist due to violated assumptions of common splicing rates and steady-state mRNA levels, leading to velocity estimation errors.

Purpose of the Study:

• To introduce scVelo, a novel method overcoming limitations in RNA velocity estimation.
• To generalize RNA velocity analysis to systems with transient cell states, crucial for development and perturbation responses.

Main Methods:

• scVelo solves the full transcriptional dynamics of splicing kinetics using a likelihood-based dynamical model.
• The method generalizes RNA velocity to transient cell states, addressing limitations of previous approaches.

Main Results:

• Application of scVelo to neurogenesis and pancreatic endocrinogenesis disentangled subpopulation kinetics.
• Inferred gene-specific rates of transcription, splicing, and degradation.
• Recovered cell positions within differentiation processes and identified potential driver genes.

Conclusions:

• scVelo provides a robust framework for analyzing RNA velocity in complex cellular systems.
• Facilitates deeper understanding of lineage decisions and gene regulation dynamics in single-cell data.