Why a file format

Scope

When a laboratory orders an engineered muscle construct, a neural culture on a microelectrode array, or a perfused tissue model from another laboratory, the criterion that decides whether the delivered batch is acceptable is usually not written down in a form that survives the transaction. It is discussed on a call, implied by a cited paper, or understood to be like the previous batch.

The consequences are: two suppliers cannot be compared because they are not quoting the same article; a batch cannot be accepted or rejected on evidence because no threshold was agreed in advance; a result cannot be reproduced by a third laboratory because the conditions that produced it were not recorded alongside the number.

What a prose specification omits

On the left is the shape of request that currently constitutes a specification for a living component. On the right is the acceptance section of the same order expressed as WWD criteria.

From: [email protected]
Subject: Re: muscle constructs, quote?

Hi, following up on the call. We need roughly 20 of the
aligned muscle strips, about 1 cm long, on the loop anchors
like the ones in the 2024 paper. Human iPSC-derived. They
need to actually contract, obviously, the last batch we got
from elsewhere were basically dead on arrival.

Ship live if you can. Happy to discuss.

AC-001  median peak twitch force  >= 500 uN
        method WWD-TWITCH-001, 37 C, 1 Hz field stimulation,
        0.5 mN preload, n = 12 random, destructive,
        one-sided test at alpha = 0.05

AC-002  twitch force CV           <= 0.20
        same 12 specimens as AC-001

AC-003  median S2D alignment      >= 0.65
        OrientationJ 2.0.5, xy plane, 5 fields, 0.25 mm2 each

PC-001  cast gel stiffness         11-13 kPa
        rheometry, tests the control axis, not the outcome

The email contains no statement that could decide a dispute. "They need to actually contract" has no threshold, statistic, sample size or assay. "The last batch were basically dead on arrival" describes a viability failure without a timepoint, a stain or a floor. "About 1 cm" is a dimension without a tolerance. "Like the ones in the paper" delegates the specification to a figure caption. Two suppliers reading it will quote different articles.

The right-hand column does not state how to make the tissue. The acceptance criteria are the contract. The process is the supplier's.

Round trip

Variance in wetware is not noise around the specification. Two constructs from the same protocol, the same passage and the same donor differ in force by amounts that matter. The package must return populated with what was measured, without overwriting what was intended.

WWD handles this with no mode flag at all. As-built is derived, not declared: a package with no measured results is as-designed, and a package with measured results is as-built in addition to still carrying its unchanged design. Every measured result references exactly one acceptance criterion by immutable UUID, and a dangling reference is a hard error rather than a warning. Appending results never mutates a criterion. A retest is a distinct result and never erases the original. Lot disposition is derived from the criteria and the results, never stored as an independent fact that somebody could set by hand.

The file is simultaneously the order form and the certificate of analysis. Every fulfilled order returns a populated file, and the accumulation of those files is a dataset mapping design to process to supplier to price to measured outcome.

WWD tags every value with a causal role as well as a provenance. Recording which quantities were set and which were observed makes each build record one row of a designed experiment rather than one row of a log. The distinction cannot be recovered after authoring.

Statistical acceptance criteria

"500 µN twitch force" omits three elements required to reproduce the judgement.

The statistic. 500 µN may be the mean, the median, the minimum, or the value of a single representative unit. For a stochastic biological product these are different promises.

The dispersion. A lot with a median of 520 µN and a coefficient of variation of 0.15 is usable. A lot with the same median and a CV of 0.60 is not: half the units fail. Location and dispersion are two criteria in WWD, never one compound statistic.

The assay. Force measured at 37 °C under 1 Hz field stimulation with 0.5 mN preload on a named fixture is not the same quantity as force measured at room temperature under paced electrode stimulation at a different preload. WWD treats a method-less functional requirement as nonconformant.

The sample size and the decision rule complete the criterion: "Median at least 500 µN, CV no greater than 0.20, n = 12, one-sided test at α = 0.05".

Non-goals

A conformant file does not mean a manufacturable design. Structural validity says the document is well-formed, the units are dimensionally consistent, the references resolve and the integrity hashes match. It says nothing about whether the tissue will survive, whether any supplier can build it, or whether the acceptance criteria are physically achievable. Passing a validator must never be described as scientific or manufacturing conformance, and the specification says so normatively.

A file format is not a solver. WWD can carry a solver-neutral model envelope, parameters, boundary conditions, run provenance and results, and a design remains completely valid without any of it. The format enforces separation: a simulated value may not be presented as a measured one, and a simulation result cannot satisfy an acceptance criterion.

Licensing and governance

Specification prose is CC BY 4.0, so it can be quoted, translated and reproduced in papers and in other standards. Schemas, validator, examples and tests are Apache-2.0, which carries an express patent grant with defensive termination. Registry vocabularies, when published, are CC0.

Those licences are granted and irrevocable. The specification, schemas, validator, conformance suite and examples are public. Nothing in the format requires a service: a .wwd file is valid, parseable and complete offline, forever, with no call to anyone's API.

Wetware World is both the format's author and its largest commercial implementer. The intent is to move WWD to neutral custody once independent implementations exist from parties with no commercial relationship to Wetware World. Until then it is a single-maintainer draft under irrevocably open licences.

Limits

There are no adopters, no external implementers and no supplier currently accepting WWD files. The feature-discovery work with laboratories that build these components has not been done. Parts of the draft are written in normative language without adoption evidence behind them.

Version 0.1 is breakable. 0.2 waits for fifty to a hundred real designs to have been pushed through 0.1. 1.0 waits until the core has survived several suppliers and several tissue types. Registry breadth is deferred so that adding a property term does not require changing the core model.

Corrections and gaps go to the participate page.