Approach

From genetics to correction.

From human genetic evidence to a gene-level correction, with a prespecified gate at each stage.

INPUTS human genetics single-cell data protein networks prior literature 01 MAP score topology GATE 02 CORRECT design intervention GATE 03 TEST falsify → validate OUTPUTS validated edit or failed model ITERATE — MODEL FAILS, REBUILD
The Loop
01

Map

Integrate human genetics, single-cell transcriptomics, protein interaction networks, and prior literature into a scored model of the disease. Each edge carries an evidence grade and can be traced back to its source.

Gate — every edge backed by reproducible evidence
02

Correct

Select the highest-leverage gene and design a gene-level correction. Precision editing targets the exact variant that drives the disease, rather than suppressing its downstream output.

Gate — correction strategy with prespecified success criteria
03

Test

Run in-silico perturbation first. If the model predicts the disease signature, move to cell and organoid validation. If it doesn't, the model is wrong and the loop resets before any expensive wet-lab spend.

Gate — perturbation reproduces disease signature
Quality Gates

Each stage has a quantitative gate agreed up front: the exact result that would disprove the model. Failure at any gate is a cheap outcome. The loop returns to Map and the model is rebuilt. The goal is to discard wrong ideas before they become expensive.

Applied To

Ileal Crohn's disease: a five-gene disease model with scored edges, ranked candidates, and a prespecified falsification test for each directed edge. The full pipeline and kill criteria are open.

→ View the preprint
Not Applicable When

The disease is polygenic without high-leverage nodes, the genetic driver is not identified, or off-target risk for the required edit is unsolved. Each fails an early gate by design.

→ See the lead programme