- RNA velocity is an important computational approach in bioinformatics that uses single-cell RNA sequencing data to estimate the direction and potential future changes of gene expression in individual cells. It provides information about cellular dynamics by examining the relationship between unspliced and spliced messenger RNA (mRNA). While conventional single-cell RNA sequencing captures gene-expression profiles at a particular sampling time, RNA velocity attempts to estimate whether a cell is moving towards a different transcriptional state. This makes it particularly useful for studying cell differentiation, development, activation, regeneration, and disease-associated cellular transitions.
- The biological foundation of RNA velocity lies in the process of RNA transcription and splicing. When a gene is transcribed, its initial RNA product is generally a precursor messenger RNA (pre-mRNA) containing both exons and introns. During RNA processing, introns are removed and exons are joined to produce mature mRNA. Unspliced RNA therefore represents newly transcribed or incompletely processed transcripts, whereas spliced RNA represents transcripts that have undergone splicing. The relative abundance of these two RNA forms provides information about the recent transcriptional activity of a gene.
- RNA velocity uses changes in unspliced and spliced RNA abundance to estimate the future direction of gene-expression changes. If a gene is being actively transcribed, its unspliced RNA abundance may increase before the corresponding mature mRNA reaches its maximum level. If transcription decreases, unspliced RNA may decline while spliced RNA remains temporarily abundant. By modelling these relationships across individual cells, computational methods estimate whether the expression of a gene is likely to increase or decrease. These gene-level estimates can then be combined to calculate a velocity vector for each cell, indicating a predicted direction of movement in a reduced-dimensional representation of the data.
- The classical mathematical framework for RNA velocity is based on a kinetic model of RNA transcription, splicing, and degradation. A simplified model can be expressed using two differential equations:
- [
\frac{du}{dt} = \alpha – \beta u
] - [
\frac{ds}{dt} = \beta u – \gamma s
] - Copyable formulas:
- du/dt = α – βu
- ds/dt = βu – γs
- Here, (u) represents the abundance of unspliced RNA, (s) represents the abundance of spliced RNA, (\alpha) is the transcription rate, (\beta) is the splicing rate, and (\gamma) is the degradation rate of mature RNA. The first equation describes the production of unspliced RNA through transcription and its removal through splicing. The second describes the production of spliced RNA from unspliced transcripts and its removal through degradation. These equations provide the foundation for estimating RNA dynamics, although practical RNA velocity methods use more detailed statistical models to account for measurement variability and gene-specific behaviour.
- The concept of RNA velocity is closely related to the idea of transcriptional states. A cell’s observed gene-expression profile represents its current molecular condition, while RNA velocity estimates the direction in which that profile may change over a short period. For example, during stem-cell differentiation, progenitor cells may show expression patterns suggesting that they are progressing towards a particular mature cell type. In immune cells, RNA velocity may reveal transitions associated with activation or changes in functional states. In developing tissues, it can help identify potential progression between intermediate and differentiated populations. However, RNA velocity estimates short-term transcriptional tendencies rather than actual chronological age, long-term developmental fate, or confirmed cell lineage.
- Several computational tools have been developed for RNA velocity analysis. Velocyto is one of the original widely used tools for quantifying unspliced and spliced RNA from single-cell sequencing data. It processes alignment information and classifies reads according to their splicing status. ScVelo, a Python-based framework, provides advanced RNA velocity modelling, visualization, and analysis. It supports stochastic and dynamical approaches that estimate transcriptional parameters and cellular transitions. Other related tools and methods include velocyto.R, kallisto-based workflows, and specialized approaches for single-cell and single-nucleus RNA sequencing. These tools can be integrated with common single-cell analysis frameworks such as Scanpy and Seurat, although RNA velocity requires suitable splicing information that may not be retained in every preprocessing workflow.
- The traditional steady-state RNA velocity model assumes that cells are sufficiently close to a transcriptional equilibrium for the relationship between unspliced and spliced RNA to be interpreted using a steady-state approximation. Under this assumption, deviations from the expected relationship can indicate whether gene expression is increasing or decreasing. However, many biological processes involve rapid transcriptional changes and do not satisfy steady-state assumptions. ScVelo therefore introduced a dynamical model that estimates transcription, splicing, and degradation rates by considering the full observed expression dynamics. This approach can identify transient transcriptional responses and more complex gene-expression patterns, although it requires sufficient data and appropriate model fitting.
- A typical RNA velocity workflow begins with single-cell RNA sequencing data that preserve information about spliced and unspliced reads. Raw sequencing reads are processed, aligned to a suitable reference genome, and classified according to their intronic and exonic mapping patterns. Tools such as STAR, Cell Ranger, and velocyto can support different stages of this process, depending on the experimental design and required output. The resulting data contain separate measurements for spliced and unspliced RNA, usually represented in count matrices. Quality control is then performed to identify low-quality cells, sequencing artefacts, and other potential sources of error. The data are normalized and analysed using a suitable RNA velocity model.
- After estimating velocity values, the results are commonly projected onto a low-dimensional representation such as PCA or UMAP. Each cell is represented by a vector or directional field that illustrates its estimated transcriptional progression. Stream plots and arrow plots can reveal potential movement between cellular populations, while velocity graphs can summarize predicted cell-to-cell transitions. These visualizations help identify possible developmental pathways and relationships between cellular states. However, the apparent direction and length of arrows depend on the modelling assumptions, data processing, and visualization method. They should not be interpreted as direct measurements of physical cell movement.
- RNA velocity is particularly useful when combined with trajectory inference and pseudotime analysis. Trajectory inference reconstructs potential paths between cellular states, while pseudotime estimates the relative position of cells along an inferred progression. RNA velocity adds directional information by estimating the short-term transcriptional changes that may occur next. Combining these approaches can help distinguish possible progression pathways, identify transitional cell populations, and investigate alternative differentiation outcomes. Tools such as Monocle 3, Slingshot, PAGA, Palantir, and tradeSeq can provide complementary information about cellular trajectories and gene-expression changes. Nevertheless, agreement between different computational methods does not replace experimental evidence of lineage relationships.
- RNA velocity can also be used to investigate gene-specific regulatory dynamics. Genes with positive velocity estimates may be increasing in expression, whereas genes with negative estimates may be decreasing, depending on the model and the definition of the velocity representation. By examining velocity patterns across genes and cells, researchers can identify transcriptional programmes associated with differentiation, activation, stress responses, or other biological transitions. Gene Ontology (GO), KEGG, Reactome, and other pathway resources can be used to interpret groups of genes associated with particular dynamic patterns. Integration with transcription-factor analysis and gene regulatory network inference may provide further insight into the mechanisms controlling cellular transitions.
- The approach has important applications in developmental biology, stem-cell research, immunology, neuroscience, cancer research, and regenerative medicine. In embryonic development, RNA velocity can help identify potential transitions from progenitor populations towards specialized cell types. In immunology, it can reveal transcriptional changes associated with immune-cell activation, maturation, and exhaustion. In cancer research, it can help investigate tumour-cell states associated with proliferation, invasion, and treatment resistance. In neuroscience, it can provide insight into neuronal and glial differentiation and changes associated with injury or disease. In regenerative medicine, it can support studies of stem-cell differentiation and tissue repair.
- RNA velocity is also valuable in animal genomics, veterinary medicine, and livestock research. Single-cell studies in cattle, sheep, goats, pigs, poultry, and other species can use RNA velocity to investigate dynamic gene-expression changes in muscle, mammary gland, reproductive organs, liver, blood, and immune tissues. In skeletal muscle, it may help characterize the progression of muscle satellite cells from quiescent states through activation, proliferation, and differentiation. In mammary tissue, it can provide information about epithelial-cell transitions during development, lactation, and tissue remodelling. In reproductive biology, it may help investigate germ-cell development and differentiation. In immune tissues, it can reveal potential transcriptional transitions associated with infection, inflammation, and vaccination. Integrating these findings with quantitative genetics, genome-wide association studies, genomic selection, and functional genomics may help connect cellular dynamics with economically important traits such as growth, milk production, fertility, disease resistance, and environmental adaptation.
- Despite its usefulness, RNA velocity has several technical and biological limitations. Accurate analysis depends on reliable classification of unspliced and spliced RNA reads, sufficient sequencing depth, and appropriate reference genome annotation. Incorrect alignment, intron retention, alternative splicing, transcriptional noise, and low gene detection can influence velocity estimates. Some genes do not follow the assumed transcriptional and splicing dynamics, while different genes may have substantially different kinetic rates. The steady-state model may be inappropriate for rapidly changing biological systems, and the dynamical model can be sensitive to data quality and model-fitting assumptions. In addition, RNA velocity does not directly establish cell lineage, developmental fate, or actual future cell behaviour.
- The interpretation of RNA velocity should therefore be supported by independent biological and experimental evidence. Time-course sequencing can help establish whether predicted transcriptional changes correspond to observed changes over time. Lineage tracing can provide evidence about actual cell ancestry, while flow cytometry, immunofluorescence, in situ hybridization, and functional experiments can validate important cellular transitions. Spatial transcriptomics can also provide information about the tissue locations of cells displaying particular velocity patterns. Combining RNA velocity with these approaches provides a stronger basis for understanding dynamic cellular processes.
- Overall, RNA velocity is a powerful computational method for studying the dynamics of gene expression in individual cells. By analysing unspliced and spliced RNA, it estimates the direction of short-term transcriptional changes and provides information that complements conventional single-cell analysis. When integrated with trajectory inference, pseudotime, gene regulatory network analysis, multi-omics, and experimental validation, RNA velocity helps researchers investigate cellular differentiation, development, activation, disease progression, and biological variation in animals and other organisms.