A safety signal surfaces — in Phase 2, Phase 3, or after approval — and a program with a valid mechanism of action gets written off.
The biology worked. Only the scaffold carried the liability.
Every existing computational tool that could help needs something you don't have at that stage: a crystal structure, a bound pose, or years of SAR data. HARMONY needs none of them.
The Problem HARMONY Solves
$200–400M. Development cost stranded when a valid drug fails on safety alone.
The biology worked. Only the scaffold carried the liability. HARMONY was built to fix that.
96%
Overall accuracy across 28 pre-registered clinical cases
12/12
Physics-only predictions correct, plus 1 honest abstention
53
Protein safety panel spanning five liability classes
94%
Of clinically actionable variant-drug pairs are in-scope
