RNA Velocity Wiki

Metabolic labeling

Properties1
tagsrna-velocity, physical-time-grounding, labeling, absolute-time

4 min read

Metabolic labeling

中文導讀

Metabolic labeling(scEU-seq、sci-fate、scNT-seq)用一段已知時長 Δt 的 nucleoside 標記 (4sU 等),把「新合成」跟「既有」RNA 分開。因為 labeling window 是已知的 real-time,它能把 absolute rate 釘住——這是少數能 break splicing-kinetics-ode 的 scale degeneracy、讓 latent time 變 metric 的外部訊號(見 physical-time-grounding)。dynamo 是吃 label 的 代表;RegVelo 則刻意不用 label,還在 driver prediction 上贏過 dynamo。

Definition

Pulse-labeling of newly synthesized RNA with a modified nucleoside (e.g. 4sU) over a known time window Δt, with chemical conversion (e.g. T→C) letting sequencing distinguish new from pre-existing transcripts per cell. Protocols: scEU-seq, sci-fate, scNT-seq.

Why it grounds physical time

Snapshot (u, s) data fixes only rate ratios — the absolute timescale of the splicing-kinetics-ode is free (see physical-time-grounding). A known Δt separates new from old RNA and thereby pins the absolute synthesis/degradation rates, converting an ordinal axis into a metric one. It is the canonical external anchor for latent-time → physical time.

Methods that use / forgo it

  • The origin used it. velocyto (2018) calibrated its extrapolation step to hours via EdU / metabolic labeling (ratio shifts detectable in 10–100 min; chromaffin extrapolation ≈ 2.5–3.8 h). Physical-time grounding was present at the start, then dropped by snapshot-only successors — see physical-time-grounding.
  • Use it — dynamo estimates absolute kinetics when labels are present.
  • Forgo it — scVelo, veloVI, RegVelo are snapshot-only. RegVelo explicitly markets recovering hematopoiesis dynamics without the metabolic labels competing methods needed — i.e. it discards the very signal that fixes the scale.

An alternative route to physical units: space, not labels

TopoVelo reaches μm/hour migration velocity without labeling — by measuring the spatial scale (real μm coordinates) and assuming the temporal scale (20 h gene cycle), then validating against live-imaging migration rates (spatial-velocity). This is metric-by-assumption, not a labeling anchor: it grounds space, not time. It shows the absolute-time problem is separable from the absolute-space problem.

A second route to metric time without labels: periodicity + a measured rate

VeloCycle (velocycle-2024, from velocyto-2018’s author GioeleLaManno) reaches metric cell-cycle time without metabolic labeling. It expresses velocity in mean-half-life units and converts to hours using a measured average mRNA half-life, exploiting the cell cycle’s known periodicity (a closed 1D manifold); the resulting period matches live-imaging microscopy + EdU — the first direct real-time validation of RNA velocity. So labeling is one anchor, not the only one: a measured absolute rate (half-life) plus a constrained, known- geometry manifold is an alternative way to break the scale degeneracy. This is the strongest metric-time case after dynamo, and the only one experimentally validated.

Caveat: labeling pins the timescale but not the growth bias

A known Δt fixes the timescale, but it does not by itself fix a second confound: cell growth. cell-growth-omission-2025 (Shah, Ming & Cleary) shows that in a growing population the estimated degradation rate absorbs the growth rate: γ* ≈ γ + λ, and the analysis targets labeling-based (4sU pulse-label / pulse-chase) designs specifically — the very methods credited here with metric time. Worse, 4sU itself slows growth during the pulse, so pulse-label and pulse-chase incorporate different growth rates, and the residual artifacts make downstream trajectories point orthogonal to / reversed from the true direction. So labeling is necessary but not sufficient for unbiased metric rates: a measured growth rate λ (or proliferation marker) is a second, independent anchor. dynamo’s “grounded with labeling” verdict therefore carries the caveat “…up to a growth bias of λ.”

NOTE: open question for FlowVelo — can flow-matching / transport carry physically interpretable time without labels, or is some absolute-time signal always required in principle? And if labels are used, does the pipeline also need a growth/dilution correction (the cell-growth-omission-2025 point)? (See physical-time-grounding open questions.)

physical-time-grounding · splicing-kinetics-ode · latent-time · spatial-velocity · dynamo · RegVelo · TopoVelo · VeloCycle · velocycle-2024 · GioeleLaManno · RNA velocity · cell-growth-omission-2025 · velocity-skepticism · FlowVelo

Graph View

Backlinks