The problem. Pseudotime (Monocle, Slingshot) orders cells along a trajectory, but the direction of travel is assumed or inferred from side information — the data itself doesn’t say which end is “before.” Can a single snapshot tell you not just where a cell sits, but which way it’s heading?

The idea. RNA velocity exploits a biological clock hidden in the data: reads spanning introns mark unspliced (newly transcribed) pre-mRNA, while spliced reads mark mature mRNA. When a gene is being switched on, unspliced runs ahead of spliced; when switched off, it lags. Modelling that ratio per gene yields a velocity vector predicting each cell’s near-future state — an arrow, not just a dot.

Why it matters. This is a genuinely clever use of information already sitting in standard RNA-seq reads (the same intron-spanning reads STAR handles). It gives trajectories a built-in arrow of time, resolving Monocle’s directionality ambiguity from first principles. It also reframes what a single-cell snapshot contains — dynamics, if you know where to look. scVelo (next) generalises it.

Verdict. Foundational and influential, if not infallible — the underlying kinetic assumptions can mislead, which later work (and scVelo’s dynamical model) addresses. Read it for the core insight: splicing state encodes direction, and it was in the data all along.