🧬 Niche and fate controls

The science

What this is. A teaching model of one tissue stem-cell niche (think of a gut crypt) and of how a progenitor cell picks a fate. It is not a medical tool: it says nothing about any treatment, dose or patient, and no part of it is medical advice.

Stem compartment (left panel). Each of the S stem cells divides at rate λ. A division is symmetric renewal (2 stem cells) with probability p, symmetric differentiation (2 progenitors) with probability q, or asymmetric (1 stem + 1 progenitor) otherwise. Niche crowding lowers the bias: (p−q)eff = (p−q)(1−S/K), so the mean follows the logistic law dS/dt = λ(p−q) S (1−S/K) (dashed white curve) while the real count is a random birth–death process, simulated exactly with Gillespie's algorithm. Small niches can lose all stem cells by chance.

Fate choice (right panel). A new progenitor starts on the ridge x = 0 of V(x) = x4/4 − x2/2 − b x and moves by overdamped Langevin dynamics dx = −V'(x) dt + σ dW. Reaching x = +1 commits it to lineage A (pink), x = −1 to lineage B (blue). The chance of A is the exit probability of the diffusion, ∫−10 e2V/σ²dx / ∫−11 e2V/σ²dx, verified below. Differentiated cells are shed at 0.3 per day.

Honest simplifications. Real stem cells have many fates, cell cycles, signals and ageing; here there is one fate axis, constant rates, no spatial niche signalling and no mutations. Landscape depth is a metaphor (Waddington, 1957). Numbers are illustrative, not measured.

Independent checks (run on load, same core as the 3D page):

running…
Tissue niche (crypt section)day 0
Waddington landscape (fate axis x)
Stem S / K-
Progenitors-
Lineage A / B-
p / q / asym-
Mean-field S-
P(A) by theory-
stemprog.AB
mean-field (dashed)
Made with 🟠 Blender