Generative AI-Designed TEAD Inhibitor Shows 40% Response Rate in Heavily

A first-in-human Phase 1 trial of ISM6331, a pill designed to block all TEAD proteins and developed using the Chemistry42 generative artificial intelligence platform, found an objective response rate of 40.0 percent among participants with heavily treated mesothelioma who could be evaluated for response at the 130 mg once-daily dose. An objective response means that a tumor shrank enough to meet the study's criteria for a response. Melina E. Marmarelis, MD, of the Abramson Cancer Center at the University of Pennsylvania, presented the findings in a Rapid Oral presentation at the European Society for Medical Oncology Congress 2026 in Madrid. The results are one of the clearest signs so far that a drug candidate designed using generative AI may work in people. Insilico Medicine, the company developing the drug and listed on the Hong Kong Stock Exchange under code 3696, announced the findings as it prepares to move the experimental treatment toward later-stage testing. The work focuses on a cancer with very few treatment options for patients whose disease has continued to progress after several treatments.

The trial, registered as NCT06566079, included 36 participants who received doses ranging from 15 mg to 130 mg. The data cutoff was August 14, 2026. These patients had already received many treatments. They had received a median of two previous lines of therapy, with numbers ranging from one to twelve. This shows how advanced their cancers were before they received the experimental drug. The study is still testing higher doses and has not yet finished choosing the Phase 1 dose. Among participants with mesothelioma who could be evaluated, the objective response rate across doses from 30 mg to 130 mg was 18.2 percent, or four of 22 people. The disease control rate was 68.2 percent, meaning that treatment either shrank tumors or kept the disease from getting worse according to the study's criteria.

Tumor responses occurred at more than one dose. Partial responses, meaning that tumors became smaller but did not necessarily disappear, were recorded at 30 mg, 50 mg, and 130 mg. This suggests that the drug may have activity at different dose levels, rather than only at the highest dose tested. At the highest dose, 130 mg once daily, two of the five participants who could be evaluated had an objective response. Their target tumors shrank by 58 percent and 36 percent. Three of the four people who responded showed signs of a response at their first scan, which took place at Week 8. Partial responses were also seen in patients who had already received five and six previous lines of therapy. Responses to a new drug are uncommon in patients who have received so many treatments.

Early results also suggested that some responses lasted over time. At the data cutoff, three of the four partial responses were still ongoing, including one that had lasted about one year. For people with mesothelioma who have used up standard treatment options, tumor shrinkage that lasts this long is an important finding. However, the study is still in an early stage and includes only a small number of patients. The company and researchers said the results support further study, rather than proving that the drug provides a clear clinical benefit. This is the appropriate way to interpret results from a Phase 1 study at this point.

The study also found signs of activity in other cancers. Target tumors became smaller in patients with heavily treated non-small cell lung cancer and head and neck squamous cell carcinoma. These patients had received seven and eight previous lines of therapy, respectively. In patients with epithelioid hemangioendothelioma and colorectal cancer, the disease remained stable for more than four months. These early findings across different cancer types fit with how the drug is expected to work, because it targets a pathway involved in many cancers. However, the company has not yet reported response rates for these other cancer types, and these observations are still exploratory.

The safety results have been an important part of the announcement. No dose-limiting toxicities, or side effects severe enough to prevent further dose increases, were reported at doses from 15 mg to 130 mg. No participant stopped treatment because of a side effect judged to be caused by the drug. Most treatment-related side effects were mild or moderate, classified as Grade 1 or Grade 2. Grade 3 treatment-related side effects occurred in 8.3 percent of participants. These included increases in the blood enzymes amylase and lipase without symptoms, tiredness, and drug-related skin inflammation. Protein in the urine, which the study identified as a side effect requiring special attention, occurred in 11.1 percent of participants. These cases were low grade and reversible. Overall, the findings support the company's view that the drug's side effects have been manageable across the doses tested so far.

The trial also produced evidence that the drug is reaching and affecting its intended target in the body. Researchers found sustained decreases in circulating CCN2, a biological marker linked to TEAD activity. This supports the conclusion that the drug is affecting the TEAD pathway in patients. The drug's terminal half-life, which is the time it takes for the amount of drug in the body to fall by half during its final stage of removal, is about 18 to 32 hours. This supports taking the drug once a day, which is practical for a pill intended to be used over a long period. These measurements are important in early cancer studies because they help show whether the drug's design leads to biological effects in people, connecting laboratory research with possible clinical responses.

ISM6331 is an experimental oral drug that blocks TEAD1, TEAD2, TEAD3, and TEAD4 without forming a permanent chemical bond with its target. It is called a pan-TEAD inhibitor because it targets all four members of the TEAD protein family rather than only some of them. The drug was designed to block gene activity controlled by TEAD proteins downstream of the Hippo pathway. This cell-signaling network can contribute to tumor growth when it does not work properly, and it is involved in several types of cancer. Insilico says it used its Chemistry42 generative AI platform to optimize the drug across several properties at once, including its ability to support once-daily dosing and its effect on CCN2 in patients. The program has received two U.S. Food and Drug Administration designations for mesothelioma: Orphan Drug Designation in June 2024 and Fast Track Designation in July 2026. The Fast Track designation was Insilico's first.

Company leaders highlighted both the difficulty of targeting TEAD and the role of artificial intelligence in developing the drug. Alex Zhavoronkov, PhD, Founder and CEO of Insilico Medicine, said the industry had considered TEAD a very difficult target for years. He said the company had now reported first-in-human results for a pan-TEAD inhibitor designed with generative AI, with a 40 percent objective response rate at 130 mg in heavily pretreated mesothelioma. He also pointed to tumor shrinkage of 58 percent and 36 percent, disease control in more than two-thirds of participants who could be evaluated, and responses by Week 8, including in people who had already received five or six lines of therapy. He noted that no dose-limiting toxicities had been reported so far and that falling levels of CCN2 in the blood suggested the drug was affecting TEAD in patients. He described the results as an example of generative chemistry producing a clinical signal. He also said Insilico sees the Hippo-TEAD pathway as relevant to both cancer and the biology of aging. The company sees possible uses across several cancers driven by this pathway and potentially in age-related diseases, in line with its strategy of developing treatments for more than one purpose. Halle Zhang, PhD, Vice President of Clinical Development for Oncology at Insilico, said the 40 percent objective response rate at 130 mg and manageable safety findings in previously treated mesothelioma support further research. She said the company plans to move to a Phase 2a randomized dose-optimization study after completing dose selection.

The next steps are to finish choosing the Phase 1 dose and, depending on the results, begin a Phase 2a randomized dose-optimization study in mesothelioma. This would be the first randomized study of ISM6331. It would help researchers determine whether the response signals seen in small groups of patients who received the drug without a comparison group lead to meaningful clinical benefits. For now, the findings come from an ongoing Phase 1 trial involving a limited number of heavily pretreated patients, and dose escalation is still underway. Even so, several findings support further study: objective responses at different doses, early and lasting tumor shrinkage in patients whose cancers had resisted previous treatments, early signs of activity in other cancers, no dose-limiting toxicities reported so far, and biological evidence that the drug is affecting its intended target. For researchers watching whether AI-designed drugs can move from computer-based design to showing activity in people, the ISM6331 results presented at ESMO 2026 offer a concrete but still early example.

Subject of Research: First-in-human Phase 1 results for ISM6331, a pan-TEAD inhibitor designed using generative AI, in heavily pretreated mesothelioma and advanced solid tumors

Article Title: Insilico Medicine reports a 40% objective response rate at 130 mg for ISM6331, a pan-TEAD inhibitor designed using generative AI, in heavily pretreated mesothelioma at ESMO 2026

Article References: Insilico Medicine reports 40% ORR at 130 mg for Pan-TEAD inhibitor ISM6331 designed using generative AI in heavily pretreated mesothelioma at ESMO 2026. Original publication (Bibliographic details unavailable.)

Image Credits: AI generated

DOI: Not provided

Keywords: ISM6331, Insilico Medicine, pan-TEAD inhibitor, mesothelioma, generative AI, Phase 1 trial, ESMO 2026, Hippo pathway, objective response rate, drug discovery, oncology, CCN2 biomarker