RNA Velocity
- Introduced by
- Gioele La Manno & Pavel Soldatov
- Year
- 2018
- Type
- Trajectory inference method
- Family
- Single-cell genomics
What does it do?
RNA velocity is a computational method that infers the future developmental state of individual cells from single-cell RNA-sequencing data.
Developed by La Manno and colleagues in 2018, RNA velocity analysis measures the direction and pace of cell state transitions by analyzing the ratio of unspliced to spliced mRNA transcripts within individual cells. This enables prediction of cell trajectories and differentiation pathways without requiring temporal sampling or manipulation, providing unique insights into cell fate decisions during development and disease.
The intuition behind the method
During transcription, mRNA is initially synthesized as unspliced precursor (pre-mRNA). Splicing removes introns, producing mature mRNA. The relative abundance of unspliced versus spliced mRNA at a gene reflects recent transcriptional activity: high unspliced/low spliced suggests recent upregulation, while low unspliced/high spliced indicates recent downregulation or continued expression of stable mRNA. RNA velocity combines this information across thousands of genes. By comparing an individual cell's transcriptional state (current gene expression) to its velocity (direction of changing expression inferred from splicing ratios), researchers can predict where the cell is headed in gene expression space. Cells with similar velocities follow similar trajectories, allowing visualization of population-level differentiation paths.
History
La Manno and colleagues introduced RNA velocity in 2018, observing that unspliced mRNA abundances in single cells provide information about transcriptional direction. They demonstrated the approach on several developmental systems (pancreatic development, neurogenesis). Bergen and colleagues (2020) extended the method with improved statistical frameworks (scVelo software), enabling better handling of transient cell states and complex kinetics. RNA velocity has rapidly become a standard post-hoc analysis in single-cell studies, enabling discovery of novel differentiation paths and validation of developmental hypotheses.
When to use it
Use RNA velocity to understand developmental trajectories, predict cell fate transitions, and validate discovered cell type relationships from clustering analysis. It is particularly powerful in developmental systems (embryogenesis, hematopoiesis) and in characterizing disease progression in tissues. Avoid RNA velocity when cells are not actively changing state (steady-state tissues with static gene expression), as velocity signals become weak. Technical consideration: droplet-based protocols (10x) require methods that infer splicing from intronic reads; full-length protocols (Smart-seq2) enable direct measurement.
Limitations and considerations
- Relies on assumption that mRNA splicing dynamics reflect future cell state changes; violates assumption for genes without ongoing transcription
- Velocity is weak in cells with low transcriptional activity (quiescent cells, mature cell types)
- Requires high-depth sequencing to accurately measure unspliced reads, especially in droplet-based protocols
- Predictions are probabilistic; cannot forecast fate of individual cells with certainty
- Sensitive to gene selection and statistical parameter choices
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Infers temporal dynamics from static snapshot data without requiring time-series experiments
- Provides cell-level resolution for developmental trajectories
- Identifies cell fate decision points and branch points in differentiation
- Validates and extends results from unsupervised clustering
- Compatible with diverse single-cell technologies and tissues
Applications
- Inferring hematopoietic differentiation trajectories from single-cell RNA-seq of bone marrow cells
- Predicting neural fate decisions during brain development by analyzing velocity in progenitor cells
- Identifying transdifferentiation pathways in reprogramming experiments where cells change fate
- Detecting early cancer cell plasticity and pre-metastatic transition signatures in single-cell tumor data
- Revealing temporal gene regulatory changes driving cell state transitions in immune cell activation
How it works
How it works
Perform single-cell RNA sequencing (e.g., 10x Genomics, Smart-seq2) on cells from a tissue or developmental stage; ensure sequencing depth is sufficient to detect both spliced and unspliced reads.
Align reads and quantify unspliced and spliced counts per gene per cell; perform standard quality control (cell filtering, gene filtering); normalize and log-transform expression data.
Compute splicing ratios (unspliced/(unspliced+spliced)) for each gene-cell combination; identify genes with robust splicing dynamics and sufficient signal.
Estimate RNA velocity by regressing splicing ratio dynamics against gene expression levels across cells; fit probabilistic models (e.g., steady-state kinetics, gamma process) to infer velocity vectors.
Visualize velocity vectors in reduced-dimensional space (e.g., UMAP, t-SNE); identify trajectory directions, terminal states, and decision points; predict cell fate choices.
Common pitfalls
- Over-interpreting velocity vectors as certain predictions rather than probabilistic estimates with uncertainty
- Applying RNA velocity to tissues with low transcriptional turnover where the method lacks power
- Assuming all genes follow simple splicing kinetics; feedback loops and transcriptional noise complicate inference
- Neglecting batch effects and technical artifacts in splicing ratio estimation
Frequently asked questions
Why does RNA velocity rely on splicing information?
Splicing dynamics reveal recent transcriptional activity. Genes being newly activated accumulate unspliced transcripts; genes being downregulated or maintaining stable expression have primarily spliced mRNA. This signal, combined with current expression levels, indicates whether gene expression is increasing or decreasing.
Can RNA velocity predict fate of single cells?
Not with certainty. RNA velocity provides probabilistic predictions based on population-level trends. Individual cells may deviate from predicted trajectories due to noise, stochasticity, or unobserved factors. Population-level predictions are more reliable than single-cell predictions.
What sequencing protocol is best for RNA velocity analysis?
Full-length protocols (Smart-seq2) are ideal because they directly capture introns and exons. Droplet-based protocols (10x Genomics) can also work if intronic reads are recovered during alignment. Protocol choice affects statistical power and technical quality of velocity estimates.
How do I validate RNA velocity predictions?
Validation approaches include: time-series experiments sampling cells at multiple timepoints (ground truth), comparison with experimental perturbation data, and consistency with known developmental biology. Software tools like CellRank integrate velocity with RNA expression and other data sources for more robust inference.
Related methods
Sources
La Manno, G., Soldatov, R., Zeisel, A., Braun, E., Hochgerner, H., Petukhov, V., & Merad, M. (2018). RNA velocity of single cells. Nature, 560(7737), 494–498.
Open DOI referenceCopied.Bergen, V., Lange, M., Peidli, S., Wolf, F. A., & Raj, B. (2020). Generalizing RNA velocity to transient cell states through smoothed differentiation of expected counts. Nature Biotechnology, 38(12), 1408–1417.
Open DOI referenceCopied.Chen, H., & Albergante, L. (2022). scVelo: RNA velocity at single-cell resolution. bioRxiv.
Open publication referenceCopied.