Where is the 'blueprint' in the C. elegans connectome?

The puzzle

C. elegans is the only animal whose entire nervous system has been mapped, connection by connection. Its wiring is famously reproducible — build a fresh worm and you get essentially the same diagram, roughly the same ~7,000 chemical synapses in roughly the same places, every time.

That reproducibility is the puzzle. The worm’s genome is tiny. It cannot possibly store a literal list of “neuron A connects to neuron B” for every connection — there simply isn’t room. So the wiring must be generated from something much more compact: a rulebook, a molecular address system, a developmental program. Somewhere there is specified information that the genome executes to build the same brain twice.

The obvious question — and the one this project chased — is where that information actually lives. Is there a hidden “wiring code” waiting to be found? Or is the wiring a by-product of simpler things, with no code at all?

This page is the honest log of the hunt. It reads as a sequence of hypotheses, each one tested with pre-registered pass/fail gates and proper statistical controls, and each one either kept or ruled out — with the reason it failed, and how that failure pointed to the next thing to try.

The investigation

kept partial ruled outtap any step to expand

What survived

After a dozen candidate “hidden codes” were tested and dissolved, three ordinary things were left carrying essentially all of the reproducible wiring. The chart below is the accounting: how far each factor gets you, and where the animal-to-animal noise floor sits.

C. elegans connectome · specification analysis

What makes the reproducible connectome reproducible?

Physical contact geometry and developmental birth-timing account for the reproducible wiring up to the animal-to-animal noise ceiling, with a small real presynaptic-transmitter bias. There is no recoverable molecular address code beneath it.

1 · Contact geometry reaches the reproducibility ceiling

The most any model can achieve is set by how well one animal's wiring predicts another's (the noise ceiling). At every synapse threshold, contact geometry meets or exceeds it — so the leftover is individual variability, not a missing code.

≥1 synapse1,257 core

geometry 0.839ceiling 0.819

≥2 synapses784 core

geometry 0.879ceiling 0.843

≥4 synapses458 core

geometry 0.905ceiling 0.880

contact-geometry AUC animal-to-animal noise ceilingaxis: 0.50 (chance) → 1.00

2 · The specification ledger

Incremental predictive power for a reproducible chemical synapse (held-out, 9,878 directed contacting pairs). Contact geometry is the dominant term; developmental arrival mediates the rest; the transmitter bias is small; lineage adds essentially nothing.

AUC 0.50 → 0.89 across the model stack

Contact geometry(+0.3245)

physical contact + area + degree — the dominant term

+ Developmental arrival(+0.0606)

birth-order timing (from L1–L3 stages, not the adult target)

+ Presynaptic NT bias(+0.0074)

small, real: Glu sources over-wire, ACh/GABA under-wire

+ Lineage grammar(+0.0018)

negligible (+0.0018) — adds nothing beyond geometry + timing

Residual

animal-to-animal variability + unmodeled — not a hidden code

3 · What it is not — the falsified hypotheses

Six candidate "hidden codes" were tested with pre-registered gates and proper nulls (degree-preserving swaps, random-gene-set controls, cross-replicate and neuron-holdout). Each was ruled out. Tap for the kill statistic.

Reading this honestly. This is a decomposition, not a discovered blueprint. Contact-dominance echoes Brittin (2021) and reproducibility echoes Witvliet (2021); the contribution here is the noise-ceiling framing, the developmental-arrival mediation, the controlled transmitter bias, and the systematic falsification of the molecular-code hypotheses. The residual is largely animal-to-animal variability — not a hidden code waiting to be found.

The honest bottom line

The reproducible worm connectome does not appear to hide a compact molecular blueprint. Wherever a hidden code could have lived — gene-similarity matching, a learned pairing grammar, a gene-regulatory generator, a lineage rulebook, a bioelectric field — a proper control dissolved it into one of three plain ingredients:

physical contact (who touches whom, and how much),
developmental timing (when each neuron is born — you cannot wire a cell before it exists), and
a small sender-side transmitter bias.

Together these explain the reliable “strong core” of the wiring right up to the animal-to-animal noise ceiling. The weak, one-off connections that vary between animals behave like tolerated noise, not like missing instructions.

Put simply: the genome seems to specify a physical-and-temporal scaffold, not an edge-by-edge wiring table. The “specified information” is distributed across contact, timing, and stabilization — not stored as a hidden code.

A note on honesty: this is a rigorous decomposition, not a discovered blueprint. That contact dominates wiring echoes Brittin et al. (2021); that the wiring is reproducible echoes Witvliet et al. (2021). The contribution here is the noise-ceiling framing, the developmental-timing mediation, the controlled transmitter bias, and — most of all — the systematic ruling-out of the hidden-code hypotheses with the right nulls. Negative results, done carefully, are results.