A cancer drug designed entirely by artificial intelligence has demonstrated striking clinical activity in one of oncology's most challenging malignancies, delivering what may be the first major validation that generative AI can create effective medicines for human patients.

Insilico Medicine's ISM6331, an oral TEAD inhibitor created using the company's Chemistry42 generative AI platform, achieved a 40% objective response rate at the 130 mg dose level in heavily pretreated mesothelioma patients during a Phase 1 first-in-human trial. The data, presented at the European Society for Medical Oncology (ESMO) Congress 2026 in Madrid, represents a significant milestone both for mesothelioma treatment and for the emerging field of AI-driven drug design.

Remarkable Activity in a Refractory Patient Population

The Phase 1 dose-escalation trial (NCT06566079) enrolled 36 participants across dose levels ranging from 15 mg to 130 mg once daily, with a data cutoff of August 14, 2026. The patient population was heavily pretreated, with a median of two prior lines of therapy and a range extending to 12 previous treatments.

Among the 22 response-evaluable mesothelioma patients treated across the 30-130 mg dose range, the overall objective response rate reached 18.2%, with four partial responses documented. At the 130 mg dose specifically, two of five evaluable patients responded, yielding the 40% response rate. Target lesion reductions at this dose level measured 58% and 36%.

The responses demonstrated clinically meaningful characteristics: three of the four responders showed evidence of tumor shrinkage at their first Week 8 scan, and three partial responses remained ongoing at data cutoff, including one lasting approximately one year. Notably, responses occurred even in patients who had received five or six prior lines of therapy, a population where treatment options are severely limited.

Beyond mesothelioma, ISM6331 showed activity signals in non-small cell lung cancer and head and neck squamous cell carcinoma in patients with seven to eight prior treatment lines, as well as stable disease lasting more than four months in epithelioid hemangioendothelioma and colorectal cancer.

Favorable Safety Profile Supports Continued Development

The safety profile proved encouraging, with no dose-limiting toxicities reported from 15 mg through 130 mg and no treatment-related adverse events leading to discontinuation. Grade 3 treatment-related adverse events occurred in just 8.3% of participants, consisting of asymptomatic amylase and lipase increases, fatigue, and drug dermatitis. Treatment-related proteinuria, identified as an adverse event of special interest, occurred in 11.1% of patients but remained low-grade and reversible. The majority of treatment-related adverse events were Grade 1-2.

ISM6331 functions as an oral, non-covalent pan-TEAD1-4 inhibitor, blocking TEAD-dependent transcription downstream of the Hippo pathway. The drug demonstrated a terminal half-life of approximately 18-32 hours, supporting once-daily dosing, and showed sustained reductions in circulating CCN2, a biomarker downstream of TEAD activity.

Validation for Generative AI Drug Discovery

The clinical results represent a potential watershed moment for generative AI in drug discovery. While multiple companies have advanced AI-designed candidates into clinical trials, ISM6331 appears to be among the first to demonstrate a clear, clinically meaningful response signal in humans. The drug was designed using Insilico's Chemistry42 platform specifically for multiparameter optimization, targeting the notoriously difficult TEAD protein family.

Insilico Medicine, a clinical-stage company listed on the Hong Kong Stock Exchange (code 3696), has received both U.S. FDA Orphan Drug Designation (June 2024) and Fast Track Designation (July 2026) for ISM6331 in mesothelioma—the company's first Fast Track program. Dose escalation continues, with Phase 1 dose selection to be completed before advancing to a Phase 2a randomized dose-optimization study in mesothelioma.

The presentation was delivered by Melina E. Marmarelis, MD, from the Abramson Cancer Center at the University of Pennsylvania, as a Rapid Oral presentation at ESMO 2026.

Sources

  1. Generative AI-Designed TEAD Inhibitor Shows 40% Response Rate in Heavily — bioengineer.org