AI-Designed TEAD Inhibitor ISM6331 Reports First-in-Human Data in Mesothelioma

A small molecule generated by an AI chemistry platform has produced its first human data against one of oncology's long-resisted transcription-factor targets. On 25 October 2026, at the European Society for Medical Oncology (ESMO) Congress in Madrid, investigators presented phase 1 results for ISM6331, an oral pan-TEAD inhibitor discovered by Insilico Medicine. The company reports tumour shrinkage in heavily pretreated mesothelioma and no dose-limiting toxicities up to 130 mg once daily. The patient numbers are very small, and so far the figures come only from the company's announcement of the Rapid Oral presentation (abstract 997, presented by Melina Marmarelis of the University of Pennsylvania's Abramson Cancer Center). Even so, the readout is a useful test case for two questions drug discoverers have been asking: can TEAD be drugged well enough to matter, and does generative design shorten the path to a clinically credible molecule?

Why TEAD has been hard to hit

The Hippo pathway is a kinase cascade that restrains cell growth. When it is active, the kinases LATS1 and LATS2 phosphorylate the co-activators YAP and TAZ and keep them out of the nucleus. When the pathway is broken, YAP and TAZ move into the nucleus and pair with the TEAD family of transcription factors (TEAD1 to TEAD4) to switch on genes for proliferation and survival. Loss of the tumour suppressor NF2, or mutations in LATS1/2, are common ways the brake fails. Hippo disruption is particularly frequent in malignant mesothelioma and has also been described in subsets of lung, head and neck and other solid tumours.

Transcription factors are notoriously difficult drug targets because they rarely have the deep, well-defined pockets that small molecules need. TEAD is a partial exception. Each TEAD protein carries a palmitate (a 16-carbon fatty acid) attached inside a central hydrophobic pocket, and this lipid modification helps stabilise TEAD and its binding to YAP and TAZ. That pocket has become the main entry point for small-molecule TEAD inhibitors. A second approach tries to block the flat protein-protein interface where YAP docks onto TEAD directly.

How ISM6331 was designed

Insilico describes ISM6331 as an orally available, non-covalent inhibitor of all four TEAD paralogues. In the company's account, its Chemistry42 generative platform started from a TEAD1 crystal structure, added structural and pharmacophore features shared by TEAD2 to TEAD4, and generated more than 6,000 candidate molecules. These were clustered into six chemical series. A first round of seven synthesised compounds produced three promising hit series. The lead was then chosen using the platform's affinity and novelty scores, followed by docking and conventional medicinal-chemistry optimisation. Insilico nominated ISM6331 as a preclinical candidate in June 2023.

The pan-TEAD design is a deliberate choice. Because the four paralogues can compensate for one another, a compound that spares some of them may leave the transcriptional programme partly intact. Blocking all four, however, raises the risk of on-target effects in normal tissues, especially the kidney, where TEAD activity appears to matter for podocyte function. The company has said preclinical work showed activity in Hippo-dysregulated tumour models, efficacy at low doses and a favourable safety and ADMET profile, and a 2024 AACR abstract reported synergy with EGFR and KRAS G12C inhibitors. These preclinical claims have not been published with full data in a peer-reviewed paper, so outside scientists cannot yet judge, for example, how selective the compound is or exactly where it binds.

"AI-designed" is easy to overstate. Generative software proposed the scaffold and helped rank candidates, but synthesis, assays, animal studies and dose selection followed the usual sequence. The reported timeline is about 19 months from candidate nomination (June 2023) to first patient dosed (January 2025), with orphan drug designation for mesothelioma in June 2024 and IND clearance in August 2024. That is quick, but not unusual for a well-run small-molecule programme, and a single compound cannot show whether the platform itself made the difference.

What the phase 1 data show

The study (NCT06566079) is a global, open-label, first-in-human trial in China and the United States. It tests ISM6331 alone in adults with advanced mesothelioma who have already received checkpoint inhibitors and platinum chemotherapy, and in other solid tumours with documented Hippo-pathway dysregulation. Its goals are safety, tolerability, pharmacokinetics, pharmacodynamics and early signs of antitumour activity, with dose escalation followed by dose optimisation.

According to Insilico's release, the data cut-off was 14 August 2026. By then, 36 participants had been enrolled at doses from 15 mg to 130 mg, with a median of two prior lines of therapy (range 1 to 12). Dose escalation was still under way. The reported efficacy figures are:

• Mesothelioma at 130 mg: objective response rate (ORR) of 40.0%, which is two partial responses among five response-evaluable patients, with target-lesion reductions of 58% and 36%.
• Mesothelioma at 30 to 130 mg combined: ORR of 18.2% (4 of 22 evaluable patients) and a disease control rate of 68.2%.
• Durability: three of the four responders had responded by the first eight-week scan, and three responses were continuing at the cut-off, one of them for about a year. Responses occurred in patients who had received five and six earlier lines of therapy.
• Other tumours: target-lesion shrinkage in single patients with non-small cell lung cancer and head and neck squamous cell carcinoma, and stable disease beyond four months in epithelioid hemangioendothelioma and colorectal cancer.

The headline 40% should be read with care. With five evaluable patients, a single extra or missing response would move the rate by 20 percentage points, and confidence intervals were not reported. The pooled figure of 18.2% across 22 patients is a more stable estimate, though still small. Without a control arm and with median survival data not yet available, the results cannot show clinical benefit. They show biological activity in a disease with few options after chemo-immunotherapy.

Safety and target engagement

Insilico reports no dose-limiting toxicities between 15 mg and 130 mg and no discontinuations due to treatment-related adverse events. Most treatment-related events were grade 1 or 2. Grade 3 events occurred in 8.3% of participants and included asymptomatic rises in amylase and lipase, fatigue and drug-related dermatitis. Treatment-related proteinuria, which the investigators tracked as an adverse event of special interest, occurred in 11.1% and was described as low grade and reversible.

Proteinuria is the safety signal to watch across the TEAD class. In a phase 1/2 study of Vivace Therapeutics' VT3989, a TEAD palmitoylation inhibitor published in Nature Medicine and presented at ESMO 2025, treatment-related increases in the urine albumin-to-creatinine ratio occurred in 31% of patients and proteinuria in 28%. Investigators handled these with dose adjustments guided by urine albumin thresholds and an intermittent schedule. ISM6331 is given once daily on a continuous schedule. It is too early to say whether its lower reported rate reflects a real difference, the doses tested so far or short follow-up.

On pharmacology, the company reports sustained falls in circulating CCN2 (also called CTGF), a classic YAP/TEAD target gene, which it presents as evidence that the drug is engaging the pathway. It also reports a terminal half-life of roughly 18 to 32 hours, consistent with once-daily dosing.

A crowded and unforgiving field

TEAD inhibition has already produced setbacks. Ikena Oncology shelved its TEAD1-selective IK-930 after the trial produced no responses. Novartis stopped work on IAG933, which blocks the YAP-TEAD interface rather than the lipid pocket. In 136 heavily pretreated patients, more than 80% with mesothelioma, IAG933 showed QTc-prolongation dose-limiting toxicities at its highest doses and an ORR of 13% among 30 pleural mesothelioma patients on continuous dosing. A company spokesperson told ApexOnco the decision reflected tolerability and activity rather than safety concerns.

VT3989 remains the reference point. In the 22 mesothelioma patients treated on the optimised intermittent schedule, the ORR was 32% and the disease control rate 86%, with median progression-free survival of about 40 weeks. ISM6331's numbers are in a similar range, but from fewer patients, an earlier cut-off and without peer review. Cross-trial comparisons at this stage say little about which agent is better.

Regulatory status

The FDA granted ISM6331 orphan drug designation for mesothelioma in June 2024. In July 2026 it granted fast track designation for adults with unresectable malignant pleural mesothelioma that progressed after anti-PD-1 therapy (with or without anti-CTLA-4) and platinum-based chemotherapy. Fast track allows more frequent interaction with the agency and possible rolling review, but neither designation establishes efficacy. ISM6331 is investigational and is not approved anywhere.

What to watch

• Dose selection: escalation beyond 130 mg is continuing, and Insilico plans a randomised phase 2a dose-optimisation study in mesothelioma if the data support it. This follows the FDA's Project Optimus emphasis on comparing doses rather than defaulting to the maximum tolerated dose.
• Kidney safety with continuous dosing: whether proteinuria stays low-grade as exposure and treatment time increase.
• Full data: a peer-reviewed publication with confidence intervals, response duration, survival outcomes and biomarker data (such as NF2 status of responders) would let independent readers judge the signal.
• Combinations: the preclinical synergy with EGFR and KRAS G12C inhibitors points to a possible role against resistance to targeted therapy, which would widen the field beyond mesothelioma.
• The AI question: a successful ISM6331 would strengthen the case that generative chemistry can open hard targets. Answering that will take outcomes across many programmes, not one early trial.

This article is for information only and is not medical advice. Treatment decisions should be made with a qualified clinician.

Sources: Insilico Medicine (PR Newswire); EurekAlert!; Targeted Oncology; Drug Target Review; The ASCO Post; ApexOnco; ClinicalTrials.gov