Laws of the laws
The same operation, run on its own output — genes into machines, machines into systems, and the laws between complexes.
Download JSON ↓What this is
The leaderboard finds laws between genes. This asset is what happens when the same operation is turned on its own output. Group the co-dependent genes and each group is a molecular machine — a protein complex the cell builds and runs as a unit. Treat each machine as a single object and ask which machines co-depend, and you get the laws between complexes. Do it once more and the whole thing collapses to eight functional systems.
None of this used a gene annotation, a pathway database, or a label. The 58 machines are recognisable complexes — SWI/SNF, SAGA, the Fanconi-anemia repair group, ATP synthase, the purine-synthesis enzymes. The 8 systems are the cell's job description: making energy, copying and repairing DNA, building proteins, processing genetic messages, holding shape, signalling, moving cargo, and dividing. It is close to the organisational chart of a living cell, drawn from raw numbers by one repeated step.
No master law
There is no single axis underneath it all. The largest pattern — roughly "how fast the cell grows" — explains only about a tenth of the variation (PC1 ≈ 11%, correlated 0.40 with overall essentiality). The cell's dependency structure is irreducibly about eight-dimensional. That is itself a finding: the machine did not flatten biology into one number, it found a small set of genuinely separate systems.
The output: laws between machines
Top 16 of 31 module-to-module couplings. Each row joins two complexes (named by their most central members). Full list, with every member gene, in the JSON.
| r | within lin. | Machine A | Machine B | sizes |
|---|---|---|---|---|
| 0.505 | 0.487 | RTCB/RPF2/TRMT61A/DDX54 | INTS14/INTS13/INTS15 | 76×3 |
| 0.459 | 0.379 | TAF5L/TADA1/TADA2B/SUPT20H | SMARCD1/BRD9/BICRA | 11×3 |
| 0.414 | 0.319 | TAF5L/TADA1/TADA2B/SUPT20H | EED/SUZ12/EZH2 | 11×3 |
| 0.401 | 0.434 | RTCB/RPF2/TRMT61A/DDX54 | TEDC1/TUBD1/TUBE1/RTTN | 76×6 |
| 0.391 | 0.408 | RTCB/RPF2/TRMT61A/DDX54 | TEN1/STN1/CTC1 | 76×3 |
| 0.373 | 0.356 | DPH2/DPH1/DPH5 | RNASEH2A/RNASEH2C/RNASEH2B | 3×3 |
| 0.371 | 0.310 | TAF5L/TADA1/TADA2B/SUPT20H | PAXIP1/PAGR1/STAG2 | 11×3 |
| 0.360 | 0.286 | NF2/AMOTL2/KIRREL1/LATS2 | SMARCD1/BRD9/BICRA | 4×3 |
| 0.359 | 0.340 | C1orf109/SPATA5L1/POP5/RPP21 | XRCC3/RAD51D/XRCC2/RAD51C | 7×4 |
| 0.358 | 0.224 | ATIC/PAICS/ADSL/GART | MED24/MED16/MED23/MED12 | 37×5 |
| 0.357 | 0.248 | TAF5L/TADA1/TADA2B/SUPT20H | TAOK1/MAP4K4/PDCD10 | 11×3 |
| 0.352 | 0.359 | RTCB/RPF2/TRMT61A/DDX54 | GET1/GET3/GET4/CAMLG | 76×4 |
| 0.351 | 0.317 | RTCB/RPF2/TRMT61A/DDX54 | UBA5/UFM1/UFL1/ODR4 | 76×6 |
| 0.348 | 0.260 | ATIC/PAICS/ADSL/GART | NF2/AMOTL2/KIRREL1/LATS2 | 37×4 |
| 0.347 | 0.306 | ARIH2/UBE2F/RNF7 | TAOK1/MAP4K4/PDCD10 | 3×3 |
| 0.344 | 0.206 | ATIC/PAICS/ADSL/GART | EED/SUZ12/EZH2 | 37×3 |
The first row is the clearest example of the method seeing across machines: an RNA-handling and ribosome-biogenesis module (carrying MRE11, RAD50, the ORC/MCM replication factors) coupled to the Integrator complex — genome maintenance meeting RNA processing, the same neighbourhood the report's METTL14–MRE11 flagship lives in.