Instead of watching NK cells patrol a physical field of tissue, this view plots every target cell directly in signal space: its horizontal position is its MHC-I (inhibitory) level, its vertical position is its stress-ligand (activating) level. The decision rule is a straight line in this plane:
S = w_A · L_activ − w_I · L_MHC-I
kill when S > θ ⇔ L_activ > L_MHC-I + θ (the boundary line, slope 1, intercept θ)
Everything above the line is the "kill zone"; everything below is the "spare zone". Downregulating MHC-I (the "missing-self" hypothesis, Kärre 1986) shifts the whole cloud of points left, past the boundary, without a single antigen ever changing — the geometry alone explains why loss of self is lethal. Raising the threshold θ slides the boundary up-left, making the NK cell more tolerant; lowering it makes it trigger-happier.
- Each dot is one target cell at its own (MHC-I, stress-ligand) coordinate — the exact same per-cell signal values as the spatial version, just plotted instead of rendered on a 3D grid.
- NK scanner agents (green rings) move through this 2D signal space itself, hopping to the nearest unengaged live cell by Euclidean distance in (MHC-I, stress-ligand) — a genuinely different notion of "distance" than physical proximity.
- Infect random patch still selects a spatially contiguous cluster (tracked internally on a hidden grid) but its effect is visible here as those points jumping across the boundary into the kill zone.
The comparison bars confirm the same signature as the spatial view: killed cells cluster at low MHC-I (left side of the plot), spared cells at high MHC-I (right side) — missing-self recognition made directly legible as geometry.