Insilico says it has 30 AI-discovered drug candidates and three programs in Phase II clinical trials
Insilico says it has 30 AI-discovered drug candidates and three programs in Phase II clinical trials
The company highlighted that figure from a Bloomberg interview with Alex Zhavoronkov and tied it to its strategy of advancing AI-designed molecules into conventional clinical development. Open registries show three Insilico Phase II entries: two for rentosertib in pulmonary fibrosis and one for ISM5411 in ulcerative colitis.
On 15 June, Insilico published an excerpt from a Bloomberg interview in which Zhavoronkov puts a number on a metric AI-biotech companies usually skate past with vague language: how many molecules have actually made it to the clinic.
"I have 30 development candidates, 3 are in Phase II, one Phase II has been completed, and we plan to move into Phase III," he said.
A development candidate is a molecule the company has chosen to push toward human trials: toxicology, manufacturing, dosing, and regulatory paperwork are all built around it. That molecule either clears the next filter or stops there, while the word "platform" remains only the name of the machine that produced it.
Insilico's official pipeline page currently shows 40+ programs, 31 preclinical candidates since 2021, and 13 IND clearances. An IND is authorization to begin a clinical trial of an experimental drug. In its April release on inhaled rentosertib, the company separately wrote that it had 30 preclinical candidates, three Phase II trials, and 13 IND clearances across fibrosis, oncology, immunology, and nervous system diseases.
ClinicalTrials.gov shows three Insilico Phase II entries: the completed Chinese study INS018_055, also known as rentosertib, in idiopathic pulmonary fibrosis; another recruiting study of the same drug in the US; and an 80-patient study in BETHESDA of ISM5411 in ulcerative colitis. The "three Phase II" figure checks out as three clinical records, two of them tied to the same drug and the same disease.
The most detailed patient-level data Insilico has published are for rentosertib. In 71 people with pulmonary fibrosis, the highest dose over 12 weeks produced a mean FVC gain of 98.4 ml, while the placebo group declined by 20.3 ml. FVC is the volume of air a person can forcefully exhale after a full inhalation; in pulmonary fibrosis, that volume falls because the tissue scars and becomes less stretchable.
In the same study, the limitations sit right next to the effect. The trial lasted 12 weeks, and the primary endpoint, meaning the main measure specified in the protocol, was safety. At high doses, some patients stopped treatment because of diarrhea and signs of liver toxicity. The company says oral rentosertib is expected to move into Phase III in the second half of 2026; that is where it will become clear whether the early signal holds up over a longer test.
For Insilico, the number 30 shifts the argument about its AI from slide decks into a queue of molecules passing through regulators, patients, and statistics. That is a harder test than another benchmark. A model can quickly suggest a target and a molecule, but a drug appears only where chemistry, manufacturing, safety, clinical design, and disease converge into a single development path.
That is why Zhavoronkov talks about a "pipeline in a product" strategy: first take a drug through a specific disease where the regulator understands the endpoint, and only then try to apply the same molecule against mechanisms of aging. It is a pragmatic workaround: aging still cannot simply be filed as a standard diagnosis, but fibrosis, inflammatory bowel disease, or tumors already have established trial rules.
The number 30 shows the scale of Insilico's bets. The probability of approvals will be revealed by late-stage trials, failures, licensing deals, and the first drugs that reach patients outside a clinical study.