Insilico Medicine's AI Discovery Offers New Hope for Hard-to-Treat Cancers

Scientists have developed a new AI-designed drug that could change how doctors treat resistant cancers.

In a recent study published in Nature Communications, Insilco Medicine launched ISM5939, a first-in-class oral inhibitor of the molecule ectonucleotide pyrophosphatase/phosphodiesterase 1 (ENPP1), which helps tumors evade the body’s defenses. By blocking ENPP1, the drug activates the body’s immune system through the Stimulator of Interferon Genes (STING) pathway—a target that has so-far been inaccessible to researchers.

This paper showcases how generative AI can not only ease the process of drug design, but also open up entirely new avenues of biological exploration—making it possible to target pathways once considered out of reach.

How Cancer Evades the Immune System

There are ~10¹⁴ cells in the human body, billions of which mutate each day. Any one of these mutations might cause uncontrollable cell growth—a prominent characteristic of tumor cells.

Under normal circumstances, the immune system acts as a vigilant gatekeeper—spotting and clearing rogue cells before they cause harm. But cancer has a way of slipping through. Some tumor cells manage to cloak themselves from detection by exploiting the body’s own regulatory systems.

One common tactic involves immune checkpoints: natural “brakes” designed to stop the immune system from going into overdrive. By pressing on those brakes, cancer cells grow—invisible to the immune system.

This led scientists to wonder if they could manipulate a patient's own immune system to fight their cancer—an idea that gave rise to the field of immunotherapy. One of the leading drug types in this field, immune checkpoint inhibitors, lift the brakes on immune cells, reactivating their “search-and-destroy” mission normally silenced by tumors.

And, while several immune checkpoint inhibitors have revolutionized the cancer treatment landscape, only ~15–25% of patients will have meaningful or durable responses. A key reason for these low percentages is the dynamic nature of tumors: they evolve, meaning that overtime they tend to stop responding to anticancer treatments.

This has sparked a push for alternative therapies: ones that can complement existing drugs or intervene when resistance emerges.

Overcoming Challenges of Targeting the STING Pathway

One promising approach involves a biological pathway called cGAS-STING, which helps immune cells detect cancer by recognizing damaged DNA. But directly targeting the pathway has brought up two major challenges:

1. Delivery issues: Direct STING activators require delivery straight into the tumor, which is not practical for tumors that have spread throughout the body.
   ‍
2. Safety concerns: Direct activation can overstimulate the immune system, causing harmful inflammation and damage to healthy tissues.

With these issues in mind, Insilico took a different approach. They turned their attention to ENPP1, an enzyme that essentially acts as a break on the STING pathway. By blocking ENPP1 with ISM5939, the researchers reasoned that removing this brake would allow the immune system to work more effectively against tumors without the harsh side effects from direct pathway activation. This approach would let natural cancer-fighting molecules accumulate where they're needed most—in and around tumors—while avoiding widespread inflammation throughout the body.

AI-Powered Drug Design: From Idea to Candidate in Months

So how did they do this?

The team used two sophisticated AI systems developed by Insilico Medicine. The first, called PandaOmics, analyzed massive amounts of biological data to identify which types of cancer would most likely respond to blocking ENPP1. The second system, Chemistry42, designed the actual drug molecule.

The AI approach identified several cancer types as prime candidates for this treatment, including breast, liver, stomach, and colorectal cancers.

Using existing ENPP1-blocking compounds as a starting point, the AI system generated thousands of new molecular designs, testing each one virtually for effectiveness, safety, and other important properties. This process, which traditionally takes years in laboratory settings, was completed in just three months.

The team's average time from starting a drug discovery program to identifying a candidate for testing is now 13 months—far shorter than the typical multi-year timeline for traditional drug development.

A Safer, Smarter Immunotherapy

In preclinical testing, ISM5939 stood out for its remarkable safety profile—a critical advantage over current immunotherapies, which can trigger dangerous inflammatory responses. Even at high doses, the new drug avoided the systemic immune overreactions that often limit other treatments.

Unlike some immunotherapies that spark a storm of inflammatory molecules across the body, ISM5939 acted with precision. It targeted ENPP1 specifically, showing minimal activity against related enzymes and suggesting a lower risk of off-target effects.

Animal studies reinforced these findings. The drug was well-tolerated, with no serious side effects observed even when administered daily over extended periods.

Amplifying the Power of Combination Therapy

ISM5939 also showed strong potential in combination with existing cancer treatments. When paired with immune checkpoint inhibitors—currently among the most effective immunotherapies—the new drug significantly boosted tumor control compared to either treatment alone.

Similar results were seen with chemotherapy. In combination with cisplatin, ISM5939 achieved better tumor suppression than either agent by itself, while maintaining a favorable safety profile.

Next Stop: Human Trials

These promising results have cleared the way for clinical testing. The U.S. Food and Drug Administration has approved Insilico Medicine’s plan to launch a Phase I trial, marking the first AI-designed ENPP1 inhibitor to reach this stage. The trial will evaluate the drug’s safety in cancer patients and help determine the appropriate dosing for future studies.

Expanding the Frontiers of Treatment

The research also highlights ISM5939’s potential to treat a broad spectrum of solid tumors. Based on molecular profiling, scientists identified eight high-priority cancers where the ENPP1 pathway appears especially active—including breast, ovarian, colorectal, stomach, esophageal, cervical, lung, and head and neck cancers.

Crucially, the team also uncovered biomarkers that may predict which patients are most likely to benefit, setting the stage for more personalized, targeted immunotherapy.

Feng Ren, Co-CEO and Chief Scientific Officer of Insilico Medicine, said:

“In this study, Insilico Medicine fully showcased the deep integration of biology, computational science, and AI-driven drug discovery and design, providing entirely new possibilities for cancer immunotherapy. “We hope the publication of the ISM5939 discovery process in Nature Communications will inspire the industry, accelerate the discovery of next-generation innovative drugs, unleash the potential of STING-targeted therapies, and bring more new options to immunotherapy.”

Concluding Remarks

The development of ISM5939 marks an exciting milestone in cancer immunotherapy, demonstrating how AI can accelerate drug discovery and unlock new treatment possibilities. By precisely targeting ENPP1 and reactivating the immune system’s natural defenses, this novel approach offers hope for patients whose tumors have become resistant to available therapies.

As AI continues to evolve, its ability to uncover hidden biological targets and design effective drugs promises to transform cancer care—bringing faster, smarter, and more personalized treatments to the clinic.