The programme

An anchor disease, and a substrate built to outgrow it.

The programme is anchored on CASK-related disorder — a neurodevelopmental condition with a few hundred known patients worldwide. CASK is the first test case, not the scope. A tool that only works for one disease has failed at the job, so every instrument takes the disease as a parameter and every reference dataset is whole-genome.

Measured

Where the substrate stands

3,275
papers appraised: 2,969
adversarial panel runs: 50
evidence corpora: 144,625
measurements extracted: 2,629
papers yielding measurements: 3,112
disease genes spanned

Derived from the research substrate on 28 August 2026 · framework 0.43.2. These are counts of our own work — papers read, panels run, measurements extracted — never a republication of any source’s content.

What it is for

Three questions the substrate is built to answer

Each is a different instrument over the same evidence base, and the third is the only one that can catch the programme being wrong.

Filter

What is worth doing

Of the proteins and pathways near the anchor gene, which are actually measured, druggable, not required by every cell that has ever been screened, and still expressed at an age when treatment could matter.

Transfer

What may be believed provisionally

For the measurements the anchor disease has no data on, which better-studied genes do — and whether their result may honestly be read across, on an explicit and inspectable warrant.

Consistency

Whether it holds together

Whether what has been measured is coherent along an ordered biological chain. This is the check that can return an answer nobody wanted, which is exactly why it is there.

Beyond the literature

Where the programme goes past reading papers

Extraction is the foundation, not the ceiling. Several layers sit on top of it — all of them hypothesis-generating, all of them upstream of the bench.

Mechanism and cross-disease connection

Where a mechanism implicated in one rare disease appears in another, which is what makes a drug studied elsewhere worth investigating here — a reason to look, never a reason to treat.

Quantitative models of the disease process

Explicit, falsifiable models with named parameters and stated scope limits, fed by the measurement substrate rather than by a single paper. A model that cannot state what would refute it is not registered.

Therapeutic strategy assessment

Approach classes scored on two gates at once — whether the mechanism can work, and whether the modality can reach the brain — because a strategy that fails either is not a strategy.

Computational molecule and binder design

A separate, licence-clean design pipeline produces candidate molecules and de-novo protein binders against implicated targets. Everything it produces is a computational prediction awaiting laboratory work.

Honesty

What is open, stated plainly

A programme whose claim is calibration owes you its own error bars first.

Human review runs well behind the machine output. That is the design — arbitration trails and never blocks — but the practical consequence is that essentially everything currently surfaced is provisional and is labelled so.
Coverage is deep in a narrow place. It is richest around the anchor disease and the neurodevelopmental genes nearest to it, and thinner as you move away. The framework reports the gap rather than filling it.
Most measurements are not yet comparable across genes. Making a value in one paper genuinely comparable to a value in another is the hard part, and a large share of the substrate has not yet cleared that bar. Those measurements are held and counted, not discarded — and not silently included either.
The generative layers have no proven forecasting skill. They propose; they do not predict. That is why every hypothesis carries the cheapest experiment that would falsify it.
Nothing here has treated anyone. This is in-silico research. It becomes meaningful for a person only after laboratory and clinical scientists have tested it — and we work with researchers who do exactly that.

Working on a rare genetic disease?

The substrate is gene-general by construction. If your disease is one of the thousands it already spans, most of the machinery points at it the day you arrive.

Request access How it works

Research aid only — not medical advice. Piper is a research and educational aid. It is not medical advice, not a diagnosis, and not a clinical determination. Its output is hypothesis-generating and may be incomplete, provisional, or wrong, including AI-generated errors. Always consult your own qualified healthcare professional before making any medical or treatment decision.