15. Killing, reverting — and the third state · Dr. Seok's Cancer-Cell Digital Twin, First Steps

In lesson 13, PDL1 knockout and CD47 knockout each did nothing at all. The reason was written in the circuit.

T_Kill* = Proliferation and not PDL1 and not CD47

There are two brakes. Release one and the other is still holding.

This is not an invented story. It is the same structure as the real problem in cancer immunotherapy, where immune-desert tumours do not respond to anti-PD1. And the answer put forward in Prof. Cho's 2025 paper was exactly to lower DDX54 and bring that wall down with it.

So what happens if we release both? Here comes the most important twist in this course.

Look at the third row again. Both numbers are 0, and yet 'cancer left' is 0 too.

Python Runs in your browser

import bn

net = bn.twin()
base = net.landscape()
cancer0 = base.states_with("cancer")

print(bn.pad("treatment", 30) + bn.pad("reversion", 12) + bn.pad("dormant", 12) + "cancer left")
print("-" * 66)
for name, p in [("PDL1 knockout (alone)",     {"PDL1": 0}),
                ("CD47 knockout (alone)",     {"CD47": 0}),
                ("PDL1 + CD47 (combined)",    {"PDL1": 0, "CD47": 0}),
                ("DDX54 knockout",            {"DDX54": 0})]:
    L = net.landscape(p)
    r = L.report(base)
    print(bn.pad(name, 30) + bn.pad(f"{r['reversion']:.0f}%", 12)
          + bn.pad(f"{r['dormant']:.0f}%", 12) + str(r["cancer_left"]))

print()
print("The combination leaves 'cancer left = 0'. And yet reversion is 0% as well.")
print("If they did not go back to normal - where on earth did the cells go?")

A new attractor has appeared — an oscillation of period 6.

Python Runs in your browser

import bn

net = bn.twin()
after = net.landscape({"PDL1": 0, "CD47": 0})
after.summary()

Growth has stopped, but DDX54 is still on. The cause is still there.

Python Runs in your browser

import bn

net = bn.twin()
cancer = net.state(DDX54=1, MDM2=1, CD47=1, PDL1=1, Proliferation=1)

print("trajectory of one cancer cell after the combined treatment")
net.trace(cancer, n=12, perturb={"PDL1": 0, "CD47": 0}, stop_when_stable=False)

print()
print("Proliferation and T_Kill take turns switching on.")
print("Immune cells keep growth suppressed but never quite eliminate it -")
print("what cancer immunology calls 'equilibrium', or dormancy.")

🎮 Three-way comparison

Apply three treatments to the same cancer cell and put the final states side by side.

The same result, 'cancer left 0', can mean three different things.

Loading…

This is relapse, reproduced inside the model.

Python Runs in your browser

import bn

net = bn.twin()

# Stop the treatment - release the perturbation from the treated state and run again
cancer = net.state(DDX54=1, MDM2=1, CD47=1, PDL1=1, Proliferation=1)

for name, p in [("combined (PDL1+CD47 KO)", {"PDL1": 0, "CD47": 0}),
                ("DDX54 knockout",          {"DDX54": 0})]:
    treated = net.run(cancer, 20, perturb=p)[-1]      # the state reached under treatment
    after_stop = net.label(net.attractor(treated))     # release the perturbation and let go
    print(bn.pad(name, 30) + "after stopping -> " + after_stop)

print()
print("Suppression comes back the moment you let go. Reversion does not.")
print("That is why reversion is worth chasing.")

🤔 Check yourself

Why do PDL1 and CD47 knockouts do nothing alone but work together?

Answer not PDL1 and not CD47 — both have to be off for it to be true. Brakes joined by AND must all be released.

🤔 Check yourself

What does the 'dormant' attractor that appears after the combined treatment mean?

🤔 Check yourself

Why must you not call the two treatments equivalent on the strength of 'cancer left = 0'?