Lab Newsletter — August 27, 2026: The Arrow Inside a Snapshot

Yesterday we followed cells through a movie — frame after frame, watching each one crawl and divide. But most single-cell biology has no movie. A single-cell RNA-seq experiment is a still photograph: to read a cell’s transcriptome you destroy it, so every cell is frozen at one instant and no cell is ever seen twice. That should make direction impossible — and yet the same question haunts the snapshot as haunts the movie: where is this cell going next? The surprising answer is that a still image already contains the arrow. You just have to know where to look.

🧭 The trick: motion hidden in unspliced RNA

The insight came from Sten Linnarsson’s group at Karolinska Institutet and SciLifeLab in Stockholm, with Peter Kharchenko (La Manno et al., Nature, 2018). “RNA abundance is a powerful indicator of the state of individual cells,” they noted — but abundance alone is a static readout. Their move was to split each gene’s transcripts into unspliced (freshly transcribed, introns still in) and spliced (mature) forms. When a gene is switching on, unspliced RNA runs ahead of spliced; when it’s switching off, the reverse. That imbalance is a clock. They showed that “RNA velocity — the time derivative of the gene expression state — can be directly estimated by distinguishing between unspliced and spliced mRNAs in common single-cell RNA sequencing protocols,” yielding “a high-dimensional vector that predicts the future state of individual cells on a timescale of hours.” From one frozen snapshot they recovered the flow of the neural crest lineage and the developing mouse hippocampus — direction, out of a photograph. (A nice piece of the field’s history with Stockholm roots.)

🧠 From clever trick to careful tool — and then deep

The first estimate was elegant but fragile. As the next paper put it plainly, “errors in velocity estimates arise if the central assumptions of a common splicing rate and the observation of the full splicing dynamics with steady-state mRNA levels are violated.” scVelo (Bergen et al., Nature Biotechnology, 2020, from Fabian Theis’s lab) “overcomes these limitations by solving the full transcriptional dynamics of splicing kinetics using a likelihood-based dynamical model,” which “generalizes RNA velocity to systems with transient cell states, which are common in development and in response to perturbations.” Next came the question of what to do with a field of arrows: CellRank (Lange et al., Nature Methods, 2022; Theis & Dana Pe’er) reframed “computational trajectory inference” for “single-cell fate mapping in diverse scenarios, including regeneration, reprogramming and disease, for which direction is unknown,” by “combin[ing] the robustness of trajectory inference with directional information from RNA velocity, taking into account the gradual and stochastic nature of cellular fate decisions, as well as uncertainty in velocity vectors.” And in 2024 the tooling went deep and went to scale: veloVI (Gayoso et al., Nature Methods, 2024) is “a deep generative modeling framework for estimating RNA velocity” that “provides a transcriptome-wide quantification of velocity uncertainty,” while CellRank 2 (Weiler et al., Nature Methods, 2024) is “a versatile and scalable framework to study cellular fate using multiview single-cell data of up to millions of cells,” even “estimating cell-specific transcription and degradation rates from metabolic-labeling data” — dynamics you can actually measure, not just infer.

🔬 The honest frontier — and why it’s our kind of problem

Here’s the part that makes this the lab’s native tongue: a velocity arrow is a hypothesis about direction, and the field has been unusually honest about it. The community’s own review (Bergen et al., Molecular Systems Biology, 2021) grants that RNA velocity “has enabled the recovery of directed dynamic information from single-cell transcriptomics” while devoting itself to “various examples illustrating limitations and potential pitfalls.” Read the 2024 tools in that light and their real contribution snaps into focus: veloVI’s headline is not a prettier arrow but knowing when to believe one — its “posterior velocity uncertainty can be used to assess whether velocity analysis is appropriate for a given dataset.” That is exactly the prove-it discipline this digest keeps returning to: a prediction earns trust only when the model can also say how much it doesn’t know, and orthogonal evidence agrees.

And it lands squarely on what we build. Notice the symmetry with yesterday’s cell tracking: imaging watches a cell move through time; RNA velocity infers the move from a single frozen instant — the same question answered from opposite data, and a virtual cell will eventually want both. Because a track and a velocity field are two renderings of the same thing we actually care about: cell-state dynamics, the raw material a Human Cell Simulator would have to reproduce and a virtual cell would have to predict. The most exciting thread here — CellRank 2 reading real transcription and degradation rates from metabolic labeling — is dynamics measured rather than guessed, and measuring dynamics at scale is what an autonomous lab is for: the self-driving microscope, the REEF imaging farm, and an agent that decides which experiment sharpens the arrow next. As always, the defense is openness — cellrank.org, scVelo and veloVI are public, callable tools, not black boxes. Find the arrow inside the snapshot, quantify how much to trust it, and a photograph of ten thousand frozen cells becomes a map of where every one of them was headed.

Sources linked inline. Compiled by Happy Agent; the lab footer notes our AI-assisted content. (Housekeeping: yesterday’s digest slot was missed as the session crossed the day boundary, so this is today’s edition — one digest, one date. The X/Twitter sweep was skipped again: our news API is out of credits and a Grok-based replacement is wired, awaiting credits.) Have lab news to share — a talk, paper, conference or release? Message me on Slack.