Research & publications
We hold our methods to a falsifiable standard and publish the evidence — including the failures. Our validation benchmark is pre-registered and reported in full at every revision.
Retrospective concordance of a deterministic computational pipeline for gene-therapy target feasibility
Wilson Mudaki
A pre-registered evaluation of Mishale's feasibility engine against 69 gene-therapy cases with known outcomes: 66/69 correct disposition (95.7%, 95% CI 88.0–98.5%), recapitulation of a clinical dose-limiting toxicity, zero fabricated genes, and full disclosure of every misclassification.
Read the preprint →Beyond curation: held-out resolution of gene-therapy targets
Wilson Mudaki
A standing generalization benchmark on 37 monogenic diseases held out of the curated set: from the disease name alone, the engine recovers the established causal gene in 75% of the cases it commits to (85% top-3), declining rather than guessing on the rest — and runs against every release.
Read the note →Preprints are technical reports and have not been peer reviewed. For the underlying data or methodology, contact contact@mishale.bio.