AI-Designed TEAD Inhibitor Enters Human Trials: Phase 1 Data for ISM6331 to Debut at ESMO

An anticancer small molecule designed with the involvement of generative artificial intelligence is set to deliver its first preliminary results from human testing. Insilico Medicine said results from a Phase 1 study of the pan-TEAD inhibitor ISM6331 have been accepted by the European Society for Medical Oncology annual meeting (ESMO 2026) and are scheduled for presentation as a rapid oral report in Madrid on October 25. For drugs still at such an early stage of development, the real test is whether humans can tolerate a dose sufficient to affect tumor biology.

ISM6331 targets TEAD transcription factors at the end of the Hippo signaling pathway. The Hippo pathway normally helps regulate cell growth, organ size, and tissue repair. When its regulation breaks down, proteins such as YAP and TAZ may activate gene programs through TEAD that promote proliferation and survival. Abnormalities in this pathway frequently occur in cancers such as malignant mesothelioma, making TEAD blockade a strategy aimed at a core tumor signal.

The study, numbered NCT06566079 and designated ISM6331-101, is an open-label, multicenter, first-in-human Phase 1 trial expected to enroll 100 patients with advanced or metastatic malignant mesothelioma and other solid tumors. ClinicalTrials.gov data show that the trial is divided into dose-escalation and dose-selection optimization portions, with study sites in the United States and China and an estimated primary completion date of August 2027.

The primary questions at this stage remain safety and tolerability, including which adverse reactions will limit dosing and whether a recommended Phase 2 dose can be selected. The study will also track the drug’s absorption and clearance in the body, its effects on the target and related biomarkers, and preliminary antitumor signals. Even if tumor shrinkage is observed, an early, nonrandomized trial is insufficient to establish efficacy. The number of patients, distribution of cancer types, and duration of responses will all affect interpretation.

The company said ISM6331 was designed with assistance from its generative chemistry platform, Chemistry42. The specific role of the AI was to explore and screen candidate molecules, not to diagnose patients or predict clinical efficacy. Whether a molecule can truly become a drug must still be determined through synthesis, experimental validation, toxicology studies, and human trials. This presentation is therefore more akin to a clinical stress test of the entire drug-design process than proof that AI has successfully developed an anticancer drug.

The company has not yet disclosed details from this set of Phase 1 data regarding patient baseline characteristics, dose levels, adverse events, or tumor responses. The October presentation most needs to clarify whether toxicity increases with exposure, whether the drug produces the expected pharmacodynamic effects in humans, and whether any early responses are concentrated in tumors with specific Hippo pathway abnormalities. These data will determine whether ISM6331 can clear the initial threshold of Phase 1 testing and will provide a more substantive basis for assessing the clinical feasibility of TEAD inhibition.