20. Wrapping up — what we built, and the real MObyDiCK · Dr. Seok's Cancer-Cell Digital Twin, First Steps

Over twenty lessons we stood a cell up inside a computer and turned the whole process of finding a therapeutic target by hand.

Genes as switches, the cell as a state vector (lesson 1)
Regulation as logic (lesson 2)
Time run forward by synchronous update (lesson 3)
Finding attractors (4) and counting basins (5)
Hanging phenotype names on the attractors (6)
Standing up the real circuit with feedback (7–9) and validating the update scheme (10)
Virtual genetic manipulation by perturbation (11–12)
Finding the target by exhaustive screening (13)
Auditing metrics and traps (14–16)
Measuring the limits of combinatorial explosion and robustness (17–18)
Handing it over as an experimental plan (19)

A conclusion we reached ourselves, by computation. Not an answer handed to us at the start.

Python Runs in your browser

import bn

net = bn.twin()
base = net.landscape()
rows = net.screen(doubles=True, quiet=True)
best = min((r for r in rows if r["harm"] == 0 and len(r["perturb"]) == 1),
           key=lambda r: (-r["reversion"], r["dormant"]))

print("twenty lessons, on one screen")
print("=" * 66)
print(f"  circuit            {net.n} genes, {len(net.edges)} regulatory relations")
print(f"  state space        {base.total}")
print(f"  attractors         {len(base.basins)} "
      f"(1 normal oscillation, 1 cancer fixed point)")
print(f"  cancer basin       {base.count('cancer')}/{base.total} "
      f"({base.count('cancer')/base.total*100:.1f}%)")
print(f"  treatments tested  {len(rows)} (single + double)")
print(f"  best single target {best['treatment']}  "
      f"reversion {best['reversion']:.0f}%, harm {best['harm']}")
print("=" * 66)

The value of a reduced model is that it is small enough to see all of.

Python Runs in your browser

import bn

net = bn.twin()
print(bn.pad("item", 22) + bn.pad("the twin in this course", 28) + "a research-scale model")
print("-" * 82)
rows = [
    ("genes",           f"{net.n}",                       "hundreds to thousands"),
    ("regulations",     f"{len(net.edges)}",              "thousands to tens of thousands"),
    ("source of edges", "invented for teaching",          "literature curation + omics data"),
    ("state space",     f"{2**net.n}",                    "2^1000 and up (no enumeration)"),
    ("attractor search","exhaustive enumeration",         "algebraic fixed points, sampling"),
    ("validation",      "shaking the rules",              "cell lines, organoids, animals"),
    ("output",          "learning",                       "papers, patents, clinical candidates"),
]
for a, b, c in rows:
    print(bn.pad(a, 22) + bn.pad(b, 28) + c)

print()
print("The scale differs by a factor of a hundred or a thousand.")
print("And every concept from lessons 1 to 19 is used unchanged -")
print("states, rules, attractors, basins, perturbation, screening, robustness.")

Congratulations on finishing all twenty lessons 🎉

Python Runs in your browser

import bn

print("where to go after this course")
print("=" * 66)
nxt = [
    ("multi-valued",     "0/1/2 instead of 0/1 - can express a partial knockdown"),
    ("stochastic",       "rules applied probabilistically (PBN) - handles noise"),
    ("differential eq.", "continuous concentration and time - precise, but many parameters"),
    ("network inference","draw the circuit automatically from data (/arac and /viper courses)"),
    ("the data side",    "reading the measured DDX54 data (/gene and /immune courses)"),
]
for k, v in nxt:
    print(f"  {bn.pad(k, 20)}{v}")

print()
print("=" * 66)
print("Well done. You can now stand a cell up inside a computer,")
print("find therapeutic targets and rank them.")

🎮 Twenty lessons, revisited

Check what each lesson taught you.

Press a blank to reveal the answer.

Loading…

🤔 Check yourself

In one sentence, what is the 'reversion' this course dealt with?

🤔 Check yourself

What is the biggest difference between our 8-gene twin and a research-scale model?

🤔 Check yourself

What does 'computation narrows the candidates; experiment answers' mean?