Why we report our experimental success rate — and how we define itPlaceholder
AI in drug discovery is only meaningful when predictions are tested. We explain the metric we track and why transparency about definitions matters.
Headline numbers in AI drug discovery are easy to produce and hard to compare. A 'hit rate' can mean very different things depending on the assay, the threshold and the starting library.
We track the share of AI-prioritised hypotheses that are confirmed in first-round wet-lab validation. Currently that figure stands at 76%. We publish the definition alongside the number, and we will refine it as our programmes mature.
Our view is simple: a model earns trust at the bench. Every result — positive or negative — is fed back into training so the next cycle is better informed.