The problem. Trajectory inference in dissociated single-cell data reconstructs a process — differentiation, activation — from a static snapshot, but it throws away where the cells were. In tissue, position often is the axis of the process (a crypt–villus gradient, a tumor margin). Can trajectory methods use spatial coordinates instead of fighting them?

The idea. This review connects trajectory/pseudotime inference to spatial data directly. It lays out how spatial context changes the problem: neighborhood information can constrain or replace purely expression-based ordering, real physical gradients can anchor a trajectory, and the usual pseudotime pitfalls (over-interpreting noise, imposing a lineage where there’s a continuum) take new forms when space is a variable. It surveys the emerging methods and the open challenges rather than crowning a winner.

Why it matters. This is a personal bridge: I did trajectory/pseudotime analysis in my AML single-cell project, so the leap to “the same analysis, but the cells keep their coordinates” is a natural extension of an existing strength rather than a cold start. For the spatial role it’s the paper that connects what I’ve already built to where the field is going.

Verdict. A review, and an early-stage one — spatial trajectory inference is genuinely less settled than clustering or deconvolution, so this is more a map of open problems than a toolbox. That’s fine; it’s the right paper for framing, and the honesty about how easily pseudotime over-interprets carries straight over from the single-cell literature (and echoes the “noisy oscillator” caution from the repressilator). Read it to connect the AML trajectory work to tissue space.