Deep quantum modeling enables new drugs for hereditary angioedema

Key takeaways:

• Current hereditary angioedema treatments have serious side effects and high costs.
• Deep quantum modeling enables exponentially greater numbers of molecular structures.
• Human trials will begin soon.

With deep quantum modeling, researchers can go beyond the millions of distinct chemical structures previously accessible to scientists and explore 10⁶⁰ possible combinations to find treatments for hereditary angioedema and other diseases.

“We use our physics breakthroughs combined with proprietary AI to discover and develop drugs that cannot be found by any other current existing methods,” Adityo Prakash, CEO of Verseon, told Healio.

Hereditary angioedema (HAE) affects at least one in every 50,000 people, which works out to at least 154,000 patients worldwide, Prakash continued.

“We have spoken to the opinion leaders in the space, and they often say that it’s underreported and underdiagnosed,” he said. “So, the actual numbers could be significantly higher.”

These patients also face a significant disease burden, Prakash said. Beyond its fluid buildup, which can be both uncomfortable and unsightly, HAE can be fatal depending on where it affects the body.

“If the swelling happens along the air passages, it can cut off your breathing, and if the restriction continues too long, the patient dies,” he said.

Diagnosis often takes 3 or more years, depending on where the patient lives, Prakash said. Patients who do not get a proper diagnosis in the first 2 decades of their life face a 25% mortality rate, and those who die of complications from HAE experience lifespans that are 31 years shorter on average.

“Clearly, it’s a serious problem,” Prakash said.

Current treatments provide prophylaxis or address acute attacks, but they often come with serious side effects, Prakash said. For example, lanadelumab-flyo (Takhzyro, Takeda), which Prakash called the leading preventive therapy, may lead to injection-site and upper respiratory infections, rash, headaches and dizziness.

C1 esterase inhibitor subcutaneous (human), or Haegarda (CSL Behring), also has been associated with injection site reactions as well as hypersensitivity, nasal pharyngitis and dizziness, among other symptoms, Prakash continued, adding that berotralstat (Orladeyo, BioCryst) has been associated with abdominal pain, vomiting, back pain, diarrhea, and QT prolongation, which adversely affects heart rhythm.

Conventional treatments also can be very expensive, Prakash said, with annual costs well exceeding $250,000 per year per patient.

By designing molecules using breakthroughs in quantum physics, Prakash said, Verseon is avoiding these side effects in the drugs that it is developing for HAE and other diseases.

“All our drugs feature novel molecular structures,” he said. “Their structures are entirely different from anything that the rest of the industry has found.”

Like lanadelumab-flyo and berotralstat, Verseon’s HAE drug is a plasma kallikrein inhibitor.

“If you design an inhibitor that binds to plasma kallikrein, it tightens up the junctions of blood vessels and prevents fluid leakage,” Prakash said. “The trouble is the agents we’re using often come with all these serious side effects and, of course, the price tag as well.”

Yet the chemical structure of Verseon’s drugs do not share any similarities with current drugs, Prakash said, with the potential to avoid any significant side effects.

“They just have a much cleaner, much better safety profile,” he said. “And we are aiming our treatments at both acute attacks and prevention.”

Verseon’s deep quantum modeling platform enabled the company to develop these chemical structures and “find drug candidates that are inaccessible to the rest of the industry,” Prakash continued.

“Artificial intelligence can’t get you there, because artificial intelligence needs a lot of training data, and then it can only predict something similar,” Prakash said.

But the number of possible drug-like chemicals is enormous, he continued, and human beings have only been able to create a miniscule fraction of all those possibilities.

“Artificial intelligence trained on the current data can’t get you into that uncharted chemical ocean,” Prakash said. “Unfortunately, most of the great medicines of the future lie in that ocean.”

Deep quantum modeling enables Verseon to design drugs based on new molecular structures, which the company then feeds into its own proprietary AI to optimize for highly desirable therapeutic properties, Prakash said.

“We actually build our own AI tools because drug discovery is an arena, especially small-molecule drug discovery, where you never have big data,” Prakash said. “You have small, sparse data.”

Typical deep learning tools do not work because of the lack of data available, he continued.

“You cannot get there by just pure brute force application of AI,” Prakash said. “If you’re going to go after designing fundamentally novel drugs that lie in that uncharted ocean of 10⁶⁰ molecules, the current droplet the industry has previously explored isn’t a big enough sample size.”

That is why Verseon had to develop its own tools, he continued.

“We have peer-reviewed published papers showing that it works far better than anything from these major companies that are doing AI,” Prakash said. “What you really need is the marriage of these deep quantum modeling breakthroughs combined with the right AI tools to actually change how drug discovery works.”

The company has completed preclinical testing of its HAE drugs in animals and is now preparing to begin human trials. It has used similar approaches in developing precision oral anticoagulants to prevent heart attacks and strokes, with one candidate completing phase 1 trials and another about to enter phase 1.

When it comes to safety, these drug candidates have passed “with flying colors, even in humans” so far, Prakash said, by preventing clots from forming without increasing bleeding risks.

“We’re seeing these promising results already across a whole range of our drug programs,” he added, including drugs for diabetes and cancer.

“I didn’t think it was possible to completely, safely stop diabetic end-organ damage like this by putting people at the earliest signs of disease on an oral drug,” he said. “We didn’t think you could develop a cancer chemo agent that doesn’t fall prey to tumor mutations.”

Prakash is optimistic about the impact that this approach will have on the future as well.

“All current AI is doing is collecting data in tiny droplets. It is not going to help you predict what’s happening way out there in the deep ocean,” Prakash said. “We’re developing drugs that you cannot find if you’re stuck in that little droplet.”