Insilico Medicine presented Phase 1 data for ISM6331 at the ESMO Congress 2026 in Madrid, and its press release reports a 40% objective response rate at the 130 mg dose in heavily pretreated mesothelioma patients. The release is dated October 25, with a page date of October 26. That is the headline. The rest of the release is where the story lives, and it is also syndicated through EurekAlert and PR Newswire.
The numbers, with their fine print
ISM6331 is a pan-TEAD inhibitor, a drug that targets TEAD proteins involved in a growth pathway some tumors rely on. Insilico says generative AI helped design the molecule. The Phase 1 trial, NCT06566079, has enrolled 36 patients across doses from 15 mg to 130 mg, and dose escalation is still underway.
At 130 mg once daily, the objective response rate was 40.0%. That equals two of five response-evaluable participants. The same passage cites target-lesion reductions of 58% and 36%, but the release does not say how many patients those figures cover. Across all mesothelioma patients dosed at 30 mg to 130 mg, the ORR is 18.2%, or four of 22 evaluable participants, with a disease control rate of 68.2% at the first tumor assessment. The company reports no dose-limiting toxicity from 15 mg through 130 mg. The data cutoff was August 14, 2026.
Regulatory signals, and what they are not
The release says ISM6331 holds U.S. FDA Orphan Drug Designation for mesothelioma, granted in June 2024, and Fast Track Designation, granted in July 2026. Those designations are real and they matter for development. They are not efficacy findings. They say the agency sees a serious condition with limited options. They do not say the drug works. Insilico's stated next step is to finish Phase 1 dose selection and then start a randomized Phase 2a dose-optimization study in mesothelioma, subject to emerging data.
What it means for solo operators and marketers
The AI angle is about the speed of the search. Generative design can propose candidate molecules faster than a traditional team working through a chemical library, but the expensive part, proving that a molecule works in people, still runs on the same clock it always did. If you are building an AI-assisted product or service, remember that template. The model shortens the front of the funnel. Validation still sets the timeline.
There is also a lesson for anyone who has ever bought a lift report. Forty percent sounds like a batting average, and technically it is one, except the sample is two hits in five at-bats. We have spent more than 25 years in digital marketing watching a small test produce a big number on a slide, and the pattern is identical here. Before you repeat the headline, ask for the denominator, the confidence interval and the control group.
Tone matters too. Mesothelioma has few good options, and the patients and families reading this story are not looking for a marketing hook. Anyone writing about it should lead with uncertainty, not enthusiasm.
Our take
Our opinion: this is a promising signal wrapped in a press release. The generative-AI origin story is interesting, but it should not earn ISM6331 extra credit in the evaluation. We judge drugs the way we judge campaigns, by outcomes at scale, in a controlled comparison, with the sample size disclosed. The randomized Phase 2a study is the event to watch, not the next conference poster. Until then, treat the 40% as a reason to pay attention, not a reason to change your roadmap. The AI-designed label will get more press than it deserves and more scrutiny than a typical molecule, which is probably the right trade. The most quoted number in this story also has the smallest denominator, and you will see that pattern again in nearly every AI announcement this year.