12. The first virtual knockout — switching DDX54 off · Dr. Seok's Cancer-Cell Digital Twin, First Steps

Now we switch one gene off and recompute the entire landscape — not one trajectory but all 256 cells.

There are three things to look at.

Is the cancer attractor still there, or gone?
Where did the cells that used to go to cancer end up?
Did any cells that were fine before get hurt (side effects)?

Do not skip the third. If you removed the cancer but turned normal cells into cancer, that is not a treatment.

The attractor count dropped from 2 to 1. The cancer valley is gone from the landscape.

Python Runs in your browser

import bn

net = bn.twin()

print("# before treatment")
before = net.landscape()
before.summary()

print("# after DDX54 knockout")
after = net.landscape({"DDX54": 0})
after.summary()

Every one went to normal, and side effects are zero. A textbook result.

Python Runs in your browser

import bn

net = bn.twin()
before = net.landscape()
after = net.landscape({"DDX54": 0})

cancer0 = before.states_with("cancer")
print(f"where did the {len(cancer0)} initial states that used to reach cancer go?")
print("-" * 54)
tally = {}
for s in cancer0:
    tally[after.fate[s]] = tally.get(after.fate[s], 0) + 1
for name, n in tally.items():
    print("  " + bn.pad(name, 12) + f"{n:>3}  ({n/len(cancer0)*100:5.1f}%)")

harmed = sum(1 for s in before.states_with("normal") if after.fate[s] == "cancer")
print()
print(f"side effects - states that were normal and became cancer: {harmed}")

As a heatmap, the order in which the dominoes fall is laid out plainly.

Python Runs in your browser

import bn
import matplotlib.pyplot as plt

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

fig, axes = plt.subplots(1, 2, figsize=(10.5, 3.6), sharey=True)
for ax, (title, p) in zip(axes, [("no treatment", None), ("DDX54 knockout", {"DDX54": 0})]):
    path = net.run(cancer, 10, perturb=p)
    M = [[s[i] for s in path] for i in range(net.n)]
    ax.imshow(M, cmap="RdYlGn_r", aspect="auto", vmin=0, vmax=1)
    ax.set_yticks(range(net.n)); ax.set_yticklabels(net.genes, fontsize=8)
    ax.set_xlabel("t"); ax.set_title(title, fontsize=11)
plt.tight_layout()

print("red = on, green = off.")
print("On the right, the top five rows switch off one after another - the dominoes.")

🎮 Knockout bench

Trajectories before and after treatment, side by side.

Try other targets too. For most of them, nothing happens at all.

Loading…

🤔 Check yourself

What must you always count as a 'side effect' when reading a knockout result?

🤔 Check yourself

What does it mean that the attractor count fell from 2 to 1 after DDX54 knockout?

🤔 Check yourself

Why did the genes switch off 'one after another' in the heatmap?