The first drug designed by generative AI just posted real Phase II lung-function data

For years the promise of AI in drug discovery has run ahead of the evidence. In June 2025 that gap narrowed. Nature Medicine published Phase IIa results for rentosertib (ISM001-055), a drug whose target and molecule were both generated using Insilico Medicine's generative AI platform. ¹

It is described as the first clinical proof-of-concept for a drug where AI drove both the choice of biological target and the design of the molecule that hits it. ²

What the trial actually found

The GENESIS-IPF trial was a double-blind, placebo-controlled study that enrolled 71 patients with idiopathic pulmonary fibrosis (IPF) across 22 sites in China. IPF is a progressive scarring of the lungs where lung function, measured as forced vital capacity (FVC), normally declines over time. ¹

• Patients on the 60 mg once-daily dose saw a mean FVC improvement of +98.4 mL.
• The placebo group declined by a mean of -20.3 mL over the same period. ¹

A swing of that size, from decline to improvement, is the kind of directional signal that justifies a larger trial. The drug was generally well tolerated, with most drug-related side effects mild or moderate; the most common were diarrhoea and abnormal liver function, each around 15 percent. ¹

Why it matters beyond one drug

The target here, a protein called TNIK, was itself surfaced by the AI approach rather than taken from existing literature. Exploratory biomarker analysis in the trial supported that mechanism. ² That is the part that should interest anyone in pharma: this was not AI optimising a known molecule against a known target, but AI proposing the biology.

Two cautions keep this honest. Phase IIa is early; it establishes a signal, not approval. And the trial was single-country with modest enrolment, so generalisability is still to be proven in larger, more diverse populations.

Still, the field now has something it lacked: a peer-reviewed, placebo-controlled human data point for an end-to-end AI-discovered drug. The debate moves from whether this can work to how reliably it can be repeated.