Dissection and integration of bursty transcriptional dynamics for complex systems.

RNA velocity estimation is a potentially powerful tool to reveal the directionality of transcriptional changes in single-cell RNA-sequencing data, but it lacks accuracy, absent advanced metabolic labeling techniques. We developed an approach, TopicVelo, that disentangles simultaneous, yet distinct, dynamics by using a probabilistic topic model, a highly interpretable form of latent space factorization, to infer cells and genes associated with individual processes, thereby capturing cellular pluripotency or multifaceted functionality. Focusing on process-associated cells and genes enables accurate estimation of process-specific velocities via a master equation for a transcriptional burst model accounting for intrinsic stochasticity. The method obtains a global transition matrix by leveraging cell topic weights to integrate process-specific signals. In challenging systems, this method accurately recovers complex transitions and terminal states, while our use of first-passage time analysis provides insights into transient transitions. These results expand the limits of RNA velocity, empowering future studies of cell fate and functional responses.

Uncovering underlying physical principles and driving forces of cell differentiation and reprogramming from single-cell transcriptomics.

Recent advances in single-cell sequencing technology have revolutionized our ability to acquire whole transcriptome data. However, uncovering the underlying transcriptional drivers and nonequilibrium driving forces of cell function directly from these data remains challenging. We address this by learning cell state vector fields from discrete single-cell RNA velocity to quantify the single-cell global nonequilibrium driving forces as landscape and flux. From single-cell data, we quantified the Waddington landscape, showing that optimal paths for differentiation and reprogramming deviate from the naively expected landscape gradient paths and may not pass through landscape saddles at finite fluctuations, challenging conventional transition state estimation of kinetic rate for cell fate decisions due to the presence of the flux. A key insight from our study is that stem/progenitor cells necessitate greater energy dissipation for rapid cell cycles and self-renewal, maintaining pluripotency. We predict optimal developmental pathways and elucidate the nucleation mechanism of cell fate decisions, with transition states as nucleation sites and pioneer genes as nucleation seeds. The concept of loop flux quantifies the contributions of each cycle flux to cell state transitions, facilitating the understanding of cell dynamics and thermodynamic cost, and providing insights into optimizing biological functions. We also infer cell-cell interactions and cell-type-specific gene regulatory networks, encompassing feedback mechanisms and interaction intensities, predicting genetic perturbation effects on cell fate decisions from single-cell omics data. Essentially, our methodology validates the landscape and flux theory, along with its associated quantifications, offering a framework for exploring the physical principles underlying cellular differentiation and reprogramming and broader biological processes through high-throughput single-cell sequencing experiments.

A single-cell atlas of mouse lung development.

Lung organogenesis requires precise timing and coordination to effect spatial organization and function of the parenchymal cells. To provide a systematic broad-based view of the mechanisms governing the dynamic alterations in parenchymal cells over crucial periods of development, we performed a single-cell RNA-sequencing time-series yielding 102,571 epithelial, endothelial and mesenchymal cells across nine time points from embryonic day 12 to postnatal day 14 in mice. Combining computational fate-likelihood prediction with RNA in situ hybridization and immunofluorescence, we explore lineage relationships during the saccular to alveolar stage transition. The utility of this publicly searchable atlas resource (www.sucrelab.org/lungcells) is exemplified by discoveries of the complexity of type 1 pneumocyte function and characterization of mesenchymal Wnt expression patterns during the saccular and alveolar stages - wherein major expansion of the gas-exchange surface occurs. We provide an integrated view of cellular dynamics in epithelial, endothelial and mesenchymal cell populations during lung organogenesis.

A single-cell gene expression atlas of human follicular aspirates: Identification of leukocyte subpopulations and their paracrine factors.

Leukocytes are in situ regulators critical for ovarian function. However, little is known about leukocyte subpopulations and their interaction with follicular cells in ovulatory follicles, especially in humans. Single-cell RNA sequencing (scRNA-seq) was performed using follicular aspirates obtained from four IVF patients and identified 13 cell groups: one granulosa cell group, one thecal cell group, 10 subsets of leukocytes, and one group of RBC/platelet. RNA velocity analyses on five granulosa cell populations predicted developmental dynamics denoting two projections of differentiation states. The cell type-specific transcriptomic profiling analyses revealed the presence of a diverse array of leukocyte-derived factors that can directly impact granulosa cell function by activating their receptors (e.g., cytokines and secretory ligands) and are involved in tissue remodeling (e.g., MMPs, ADAMs, ADAMTSs, and TIMPs) and angiogenesis (e.g., VEGFs, PGF, FGF, IGF, and THBS1) in ovulatory follicles. Consistent with the findings from the scRNA-seq data, the leukocyte-specific expression of CD68, IL1B, and MMP9 was verified in follicle tissues collected before and at defined hours after hCG administration from regularly cycling women. Collectively, this study demonstrates that this data can be used as an invaluable resource for identifying important leukocyte-derived factors that promote follicular cell function, thereby facilitating ovulation and luteinization in women.

Representation learning of RNA velocity reveals robust cell transitions.

RNA velocity is a promising technique for quantifying cellular transitions from single-cell transcriptome experiments and revealing transient cellular dynamics among a heterogeneous cell population. However, the cell transitions estimated from high-dimensional RNA velocity are often unstable or inaccurate, partly due to the high technical noise and less informative projection. Here, we present Velocity Autoencoder (VeloAE), a tailored representation learning method, to learn a low-dimensional representation of RNA velocity on which cellular transitions can be robustly estimated. On various experimental datasets, we show that VeloAE can both accurately identify stimulation dynamics in time-series designs and effectively capture expected cellular differentiation in different biological systems. VeloAE, therefore, enhances the usefulness of RNA velocity for studying a wide range of biological processes.

A high-resolution transcriptomic atlas depicting nitrogen fixation and nodule development in soybean.

Although root nodules are essential for biological nitrogen fixation in legumes, the cell types and molecular regulatory mechanisms contributing to nodule development and nitrogen fixation in determinate nodule legumes, such as soybean (Glycine max), remain incompletely understood. Here, we generated a single-nucleus resolution transcriptomic atlas of soybean roots and nodules at 14 days post inoculation (dpi) and annotated 17 major cell types, including six that are specific to nodules. We identified the specific cell types responsible for each step in the ureides synthesis pathway, which enables spatial compartmentalization of biochemical reactions during soybean nitrogen fixation. By utilizing RNA velocity analysis, we reconstructed the differentiation dynamics of soybean nodules, which differs from those of indeterminate nodules in Medicago truncatula. Moreover, we identified several putative regulators of soybean nodulation and two of these genes, GmbHLH93 and GmSCL1, were as-yet uncharacterized in soybean. Overexpression of each gene in soybean hairy root systems validated their respective roles in nodulation. Notably, enrichment for cytokinin-related genes in soybean nodules led to identification of the cytokinin receptor, GmCRE1, as a prominent component of the nodulation pathway. GmCRE1 knockout in soybean resulted in a striking nodule phenotype with decreased nitrogen fixation zone and depletion of leghemoglobins, accompanied by downregulation of nodule-specific gene expression, as well as almost complete abrogation of biological nitrogen fixation. In summary, this study provides a comprehensive perspective of the cellular landscape during soybean nodulation, shedding light on the underlying metabolic and developmental mechanisms of soybean nodule formation.

RNA velocity-current challenges and future perspectives.

RNA velocity has enabled the recovery of directed dynamic information from single-cell transcriptomics by connecting measurements to the underlying kinetics of gene expression. This approach has opened up new ways of studying cellular dynamics. Here, we review the current state of RNA velocity modeling approaches, discuss various examples illustrating limitations and potential pitfalls, and provide guidance on how the ensuing challenges may be addressed. We then outline future directions on how to generalize the concept of RNA velocity to a wider variety of biological systems and modalities.

Spatial transcriptome profiling by MERFISH reveals subcellular RNA compartmentalization and cell cycle-dependent gene expression.

The expression profiles and spatial distributions of RNAs regulate many cellular functions. Image-based transcriptomic approaches provide powerful means to measure both expression and spatial information of RNAs in individual cells within their native environment. Among these approaches, multiplexed error-robust fluorescence in situ hybridization (MERFISH) has achieved spatially resolved RNA quantification at transcriptome scale by massively multiplexing single-molecule FISH measurements. Here, we increased the gene throughput of MERFISH and demonstrated simultaneous measurements of RNA transcripts from ∼10,000 genes in individual cells with ∼80% detection efficiency and ∼4% misidentification rate. We combined MERFISH with cellular structure imaging to determine subcellular compartmentalization of RNAs. We validated this approach by showing enrichment of secretome transcripts at the endoplasmic reticulum, and further revealed enrichment of long noncoding RNAs, RNAs with retained introns, and a subgroup of protein-coding mRNAs in the cell nucleus. Leveraging spatially resolved RNA profiling, we developed an approach to determine RNA velocity in situ using the balance of nuclear versus cytoplasmic RNA counts. We applied this approach to infer pseudotime ordering of cells and identified cells at different cell-cycle states, revealing ∼1,600 genes with putative cell cycle-dependent expression and a gradual transcription profile change as cells progress through cell-cycle stages. Our analysis further revealed cell cycle-dependent and cell cycle-independent spatial heterogeneity of transcriptionally distinct cells. We envision that the ability to perform spatially resolved, genome-wide RNA profiling with high detection efficiency and accuracy by MERFISH could help address a wide array of questions ranging from the regulation of gene expression in cells to the development of cell fate and organization in tissues.

Cascade Forest-Based Model for Prediction of RNA Velocity.

In recent years, single-cell RNA sequencing technology (scRNA-seq) has developed rapidly and has been widely used in biological and medical research, such as in expression heterogeneity and transcriptome dynamics of single cells. The investigation of RNA velocity is a new topic in the study of cellular dynamics using single-cell RNA sequencing data. It can recover directional dynamic information from single-cell transcriptomics by linking measurements to the underlying dynamics of gene expression. Predicting the RNA velocity vector of each cell based on its gene expression data and formulating RNA velocity prediction as a classification problem is a new research direction. In this paper, we develop a cascade forest model to predict RNA velocity. Compared with other popular ensemble classifiers, such as XGBoost, RandomForest, LightGBM, NGBoost, and TabNet, it performs better in predicting RNA velocity. This paper provides guidance for researchers in selecting and applying appropriate classification tools in their analytical work and suggests some possible directions for future improvement of classification tools.

Velo-Predictor: an ensemble learning pipeline for RNA velocity prediction.

BACKGROUND: RNA velocity is a novel and powerful concept which enables the inference of dynamical cell state changes from seemingly static single-cell RNA sequencing (scRNA-seq) data. However, accurate estimation of RNA velocity is still a challenging problem, and the underlying kinetic mechanisms of transcriptional and splicing regulations are not fully clear. Moreover, scRNA-seq data tend to be sparse compared with possible cell states, and a given dataset of estimated RNA velocities needs imputation for some cell states not yet covered. RESULTS: We formulate RNA velocity prediction as a supervised learning problem of classification for the first time, where a cell state space is divided into equal-sized segments by directions as classes, and the estimated RNA velocity vectors are considered as ground truth. We propose Velo-Predictor, an ensemble learning pipeline for predicting RNA velocities from scRNA-seq data. We test different models on two real datasets, Velo-Predictor exhibits good performance, especially when XGBoost was used as the base predictor. Parameter analysis and visualization also show that the method is robust and able to make biologically meaningful predictions. CONCLUSION: The accurate result shows that Velo-Predictor can effectively simplify the procedure by learning a predictive model from gene expression data, which could help to construct a continous landscape and give biologists an intuitive picture about the trend of cellular dynamics.

Transcriptional dynamics of pluripotent stem cell-derived endothelial cell differentiation revealed by single-cell RNA sequencing.

AIMS: Pluripotent stem cell-derived endothelial cell products possess therapeutic potential in ischaemic vascular disease. However, the factors that drive endothelial differentiation from pluripotency and cellular specification are largely unknown. The aims of this study were to use single-cell RNA sequencing (scRNA-seq) to map the transcriptional landscape and cellular dynamics of directed differentiation of human embryonic stem cell-derived endothelial cells (hESC-EC) and to compare these cells to mature endothelial cells from diverse vascular beds. METHODS AND RESULTS: A highly efficient directed 8-day differentiation protocol was used to generate a hESC-derived endothelial cell product (hESC-ECP), in which 66% of cells co-expressed CD31 and CD144. We observed largely homogeneous hESC and mesodermal populations at Days 0 and 4, respectively, followed by a rapid emergence of distinct endothelial and mesenchymal populations. Pseudotime trajectory identified transcriptional signatures of endothelial commitment and maturation during the differentiation process. Concordance in transcriptional signatures was verified by scRNA-seq analysis using both a second hESC line RC11, and an alternative hESC-EC differentiation protocol. In total, 105 727 cells were subjected to scRNA-seq analysis. Global transcriptional comparison revealed a transcriptional architecture of hESC-EC that differs from freshly isolated and cultured human endothelial cells and from organ-specific endothelial cells. CONCLUSION: A transcriptional bifurcation into endothelial and mesenchymal lineages was identified, as well as novel transcriptional signatures underpinning commitment and maturation. The transcriptional architecture of hESC-ECP was distinct from mature and foetal human EC.

Reconstructing developmental trajectories using latent dynamical systems and time-resolved transcriptomics.

The snapshot nature of single-cell transcriptomics presents a challenge for studying the dynamics of cell fate decisions. Metabolic labeling and splicing can provide temporal information at single-cell level, but current methods have limitations. Here, we present a framework that overcomes these limitations: experimentally, we developed sci-FATE2, an optimized method for metabolic labeling with increased data quality, which we used to profile 45,000 embryonic stem (ES) cells differentiating into neural tube identities. Computationally, we developed a two-stage framework for dynamical modeling: VelvetVAE, a variational autoencoder (VAE) for velocity inference that outperforms all other tools tested, and VelvetSDE, a neural stochastic differential equation (nSDE) framework for simulating trajectory distributions. These recapitulate underlying dataset distributions and capture features such as decision boundaries between alternative fates and fate-specific gene expression. These methods recast single-cell analyses from descriptions of observed data to models of the dynamics that generated them, providing a framework for investigating developmental fate decisions.

DeepVelo: deep learning extends RNA velocity to multi-lineage systems with cell-specific kinetics.

Existing RNA velocity estimation methods strongly rely on predefined dynamics and cell-agnostic constant transcriptional kinetic rates, assumptions often violated in complex and heterogeneous single-cell RNA sequencing (scRNA-seq) data. Using a graph convolution network, DeepVelo overcomes these limitations by generalizing RNA velocity to cell populations containing time-dependent kinetics and multiple lineages. DeepVelo infers time-varying cellular rates of transcription, splicing, and degradation, recovers each cell's stage in the differentiation process, and detects functionally relevant driver genes regulating these processes. Application to various developmental and pathogenic processes demonstrates DeepVelo's capacity to study complex differentiation and lineage decision events in heterogeneous scRNA-seq data.

Single-nucleus RNA velocity reveals critical synaptic and cell-cycle dysregulations in neuropathologically confirmed Alzheimer's disease.

Typical differential single-nucleus gene expression (snRNA-seq) analyses in Alzheimer's disease (AD) provide fixed snapshots of cellular alterations, making the accurate detection of temporal cell changes challenging. To characterize the dynamic cellular and transcriptomic differences in AD neuropathology, we apply the novel concept of RNA velocity to the study of single-nucleus RNA from the cortex of 60 subjects with varied levels of AD pathology. RNA velocity captures the rate of change of gene expression by comparing intronic and exonic sequence counts. We performed differential analyses to find the significant genes driving both cell type-specific RNA velocity and expression differences in AD, extensively compared these two transcriptomic metrics, and clarified their associations with multiple neuropathologic traits. The results were cross-validated in an independent dataset. Comparison of AD pathology-associated RNA velocity with parallel gene expression differences reveals sets of genes and molecular pathways that underlie the dynamic and static regimes of cell type-specific dysregulations underlying the disease. Differential RNA velocity and its linked progressive neuropathology point to significant dysregulations in synaptic organization and cell development across cell types. Notably, most of the genes underlying this synaptic dysregulation showed increased RNA velocity in AD subjects compared to controls. Accelerated cell changes were also observed in the AD subjects, suggesting that the precocious depletion of precursor cell pools might be associated with neurodegeneration. Overall, this study uncovers active molecular drivers of the spatiotemporal alterations in AD and offers novel insights towards gene- and cell-centric therapeutic strategies accounting for dynamic cell perturbations and synaptic disruptions.

Causal gene regulatory analysis with RNA velocity reveals an interplay between slow and fast transcription factors.

Single-cell expression dynamics, from differentiation trajectories or RNA velocity, have the potential to reveal causal links between transcription factors (TFs) and their target genes in gene regulatory networks (GRNs). However, existing methods either overlook these expression dynamics or necessitate that cells be ordered along a linear pseudotemporal axis, which is incompatible with branching trajectories. We introduce Velorama, an approach to causal GRN inference that represents single-cell differentiation dynamics as a directed acyclic graph of cells, constructed from pseudotime or RNA velocity measurements. Additionally, Velorama enables the estimation of the speed at which TFs influence target genes. Applying Velorama, we uncover evidence that the speed of a TF's interactions is tied to its regulatory function. For human corticogenesis, we find that slow TFs are linked to gliomas, while fast TFs are associated with neuropsychiatric diseases. We expect Velorama to become a critical part of the RNA velocity toolkit for investigating the causal drivers of differentiation and disease.

Forseti: A mechanistic and predictive model of the splicing status of scRNA-seq reads.

Motivation: Short-read single-cell RNA-sequencing (scRNA-seq) has been used to study cellular heterogeneity, cellular fate, and transcriptional dynamics. Modeling splicing dynamics in scRNA-seq data is challenging, with inherent difficulty in even the seemingly straightforward task of elucidating the splicing status of the molecules from which sequenced fragments are drawn. This difficulty arises, in part, from the limited read length and positional biases, which substantially reduce the specificity of the sequenced fragments. As a result, the splicing status of many reads in scRNA-seq is ambiguous because of a lack of definitive evidence. We are therefore in need of methods that can recover the splicing status of ambiguous reads which, in turn, can lead to more accuracy and confidence in downstream analyses. Results: We develop Forseti, a predictive model to probabilistically assign a splicing status to scRNA-seq reads. Our model has two key components. First, we train a binding affinity model to assign a probability that a given transcriptomic site is used in fragment generation. Second, we fit a robust fragment length distribution model that generalizes well across datasets deriving from different species and tissue types. Forseti combines these two trained models to predict the splicing status of the molecule of origin of reads by scoring putative fragments that associate each alignment of sequenced reads with proximate potential priming sites. Using both simulated and experimental data, we show that our model can precisely predict the splicing status of reads and identify the true gene origin of multi-gene mapped reads. Availability: Forseti and the code used for producing the results are available at https://github.com/COMBINE-lab/forseti under a BSD 3-clause license.

Single-Cell Transcriptome Analysis Identified Core Genes and Transcription Factors in Mesenchymal Cell Differentiation during Liver Cirrhosis.

BACKGROUND: Mesenchymal cells, including hepatic stellate cells (HSCs), fibroblasts (FBs), myofibroblasts (MFBs), and vascular smooth muscle cells (VSMCs), are the main cells that affect liver fibrosis and play crucial roles in maintaining tissue homeostasis. The dynamic evolution of mesenchymal cells is very important but remains to be explored for researching the reversible mechanism of hepatic fibrosis and its evolution mechanism of hepatic fibrosis to cirrhosis. METHODS: Here, we analysed the transcriptomes of more than 50,000 human single cells from three cirrhotic and three healthy liver tissue samples and the mouse hepatic mesenchymal cells of two healthy and two fibrotic livers to reconstruct the evolutionary trajectory of hepatic mesenchymal cells from a healthy to a cirrhotic state, and a subsequent integrative analysis of bulk RNA sequencing (RNA-seq) data of HSCs from quiescent to active (using transforming growth factor ß1 (TGF-ß1) to stimulate LX-2) to inactive states. RESULTS: We identified core genes and transcription factors (TFs) involved in mesenchymal cell differentiation. In healthy human and mouse livers, the expression of NR1H4 and members of the ZEB families (ZEB1 and ZEB2) changed significantly with the differentiation of FB into HSC and VSMC. In cirrhotic human livers, VSMCs transformed into HSCs with downregulation of MYH11, ACTA2, and JUNB and upregulation of PDGFRB, RGS5, IGFBP5, CD36, A2M, SOX5, and MEF2C. Following HSCs differentiation into MFBs with the upregulation of COL1A1, TIMP1, and NR1H4, a small number of MFBs reverted to inactivated HSCs (iHSCs). The differentiation trajectory of mouse hepatic mesenchymal cells was similar to that in humans; however, the evolution trajectory and proportion of cell subpopulations that reverted from MFBs to iHSCs suggest that the mouse model may not accurately reflect disease progression and outcome in humans. CONCLUSIONS: Our analysis elucidates primary genes and TFs involved in mesenchymal cell differentiation during liver fibrosis using scRNA-seq data, and demonstrated the core genes and TFs in process of HSC activation to MFB and MFB reversal to iHSC using bulk RNA-seq data of human fibrosis induced by TGF-ß1. Furthermore, our findings suggest promising targets for the treatment of liver fibrosis and provide valuable insights into the molecular mechanisms underlying its onset and progression.

snRNA-seq of human cutaneous neurofibromas before and after selumetinib treatment implicates role of altered Schwann cell states, inter-cellular signaling, and extracellular matrix in treatment response.

Neurofibromatosis Type 1 (NF1) is caused by loss of function variants in the NF1 gene. Most patients with NF1 develop skin lesions called cutaneous neurofibromas (cNFs). Currently the only approved therapeutic for NF1 is selumetinib, a mitogen -activated protein kinase (MEK) inhibitor. The purpose of this study was to analyze the transcriptome of cNF tumors before and on selumetinib treatment to understand both tumor composition and response. We obtained biopsy sets of tumors both pre- and on- selumetinib treatment from the same individuals and were able to collect sets from four separate individuals. We sequenced mRNA from 5844 nuclei and identified 30,442 genes in the untreated group and sequenced 5701 nuclei and identified 30,127 genes in the selumetinib treated group. We identified and quantified distinct populations of cells (Schwann cells, fibroblasts, pericytes, myeloid cells, melanocytes, keratinocytes, and two populations of endothelial cells). While we anticipated that cell proportions might change with treatment, we did not identify any one cell population that changed significantly, likely due to an inherent level of variability between tumors. We also evaluated differential gene expression based on drug treatment in each cell type. Ingenuity pathway analysis (IPA) was also used to identify pathways that differ on treatment. As anticipated, we identified a significant decrease in ERK/MAPK signaling in cells including Schwann cells but most specifically in myeloid cells. Interestingly, there is a significant decrease in opioid signaling in myeloid and endothelial cells; this downward trend is also observed in Schwann cells and fibroblasts. Cell communication was assessed by RNA velocity, Scriabin, and CellChat analyses which indicated that Schwann cells and fibroblasts have dramatically altered cell states defined by specific gene expression signatures following treatment (RNA velocity). There are dramatic changes in receptor-ligand pairs following treatment (Scriabin), and robust intercellular signaling between virtually all cell types associated with extracellular matrix (ECM) pathways (Collagen, Laminin, Fibronectin, and Nectin) is downregulated after treatment. These response specific gene signatures and interaction pathways could provide clues for understanding treatment outcomes or inform future therapies.

DeepKINET: a deep generative model for estimating single-cell RNA splicing and degradation rates.

Messenger RNA splicing and degradation are critical for gene expression regulation, the abnormality of which leads to diseases. Previous methods for estimating kinetic rates have limitations, assuming uniform rates across cells. DeepKINET is a deep generative model that estimates splicing and degradation rates at single-cell resolution from scRNA-seq data. DeepKINET outperforms existing methods on simulated and metabolic labeling datasets. Applied to forebrain and breast cancer data, it identifies RNA-binding proteins responsible for kinetic rate diversity. DeepKINET also analyzes the effects of splicing factor mutations on target genes in erythroid lineage cells. DeepKINET effectively reveals cellular heterogeneity in post-transcriptional regulation.

Pumping the brakes on RNA velocity by understanding and interpreting RNA velocity estimates.

BACKGROUND: RNA velocity analysis of single cells offers the potential to predict temporal dynamics from gene expression. In many systems, RNA velocity has been observed to produce a vector field that qualitatively reflects known features of the system. However, the limitations of RNA velocity estimates are still not well understood. RESULTS: We analyze the impact of different steps in the RNA velocity workflow on direction and speed. We consider both high-dimensional velocity estimates and low-dimensional velocity vector fields mapped onto an embedding. We conclude the transition probability method for mapping velocity estimates onto an embedding is effectively interpolating in the embedding space. Our findings reveal a significant dependence of the RNA velocity workflow on smoothing via the k-nearest-neighbors (k-NN) graph of the observed data. This reliance results in considerable estimation errors for both direction and speed in both high- and low-dimensional settings when the k-NN graph fails to accurately represent the true data structure; this is an unknown feature of real data. RNA velocity performs poorly at estimating speed in both low- and high-dimensional spaces, except in very low noise settings. We introduce a novel quality measure that can identify when RNA velocity should not be used. CONCLUSIONS: Our findings emphasize the importance of choices in the RNA velocity workflow and highlight critical limitations of data analysis. We advise against over-interpreting expression dynamics using RNA velocity, particularly in terms of speed. Finally, we emphasize that the use of RNA velocity in assessing the correctness of a low-dimensional embedding is circular.