With a nudge from AI, ketamine emerges as potential rare disease drug

In the seven difficult years since their son Mateo was diagnosed with a rare disease, Victoria Malvagno and Frank Solorzano have been waiting for medicine to catch up with their lives.

Doctors tested Mateo for hundreds of conditions before they finally determined he was one of only a few hundred people in the world with a neurodevelopmental condition called ADNP syndrome. Even with a diagnosis in hand, just getting through each day has been a full-time job. Mateo communicates mostly nonverbally, and his parents must constantly be on alert to make sure he doesn’t hurt himself. For years, they have muddled through managing his many symptoms piecemeal, because there’s no treatment approved for the rare genetic disease.

But in 2019, a breakthrough came — not from a doctor’s office, or a drug company, or another battery of tests. It came from an artificial intelligence system. The tool, called mediKanren, scanned millions of biomedical abstracts hunting for relationships between existing compounds and the gene involved in the disease.

advertisement

In seconds, it pinpointed an unexpected target: ketamine.

Ketamine’s history as a recreational drug might make it seem a dubious candidate to treat a rare disease in children. But it has long been approved as an anesthesia drug in kids and adults, and researchers were finalizing its approval as a treatment for major depression. The AI suggested it could be repurposed to fight a new foe — a strategy researchers hope will bear fruit for many untreated rare diseases.

This combination of artificial intelligence and a repurposed drug has enabled researchers to accelerate their usual timelines. In 2020, just a year and half after the system pinged on its result, the first clinical trial for ADNP syndrome was underway.

advertisement

Last month, there was a hint of progress, when the trial reported preliminary safety results in 10 children, including Mateo, now a rambunctious 13-year-old who loves listening to music and splashing in sprinklers near his home in Brooklyn, N.Y.

Yet, cutting edge as it may be, researchers and families are mindful that technology can only do so much to speed up their fervent search for a treatment. Clinical trials are a knotty, costly business, and they’re especially difficult to conduct for rare diseases. Drug companies rarely invest in testing treatments with such a small potential market, despite financial incentives and expedited pathways from the Food and Drug Administration.

So as the pace of science slows back down, Mateo’s family and others affected by ADNP syndrome are stuck waiting again.

“For us as parents,” said Malvagno, “it can’t happen fast enough.”

The ketamine research that mediKanren found was years old, but nobody had advocated for the drug’s potential as a treatment for ADNP syndrome — until it landed in the inbox of a neurosurgeon named Matt Davis.

In February 2019, Davis was nearing the end of his residency at the University of Alabama at Birmingham, catching a quick moment between cases. “I was waiting for a patient to wake up from surgery and looking on my phone, and I got this email from Matt,” said Davis.

The other Matt was Matt Might, director at UAB’s Hugh Kaul Precision Medicine Institute. The two were connected not just by their university, but by their experience as fathers of children with rare disease. Davis’ son Benjamin had been diagnosed with ADNP syndrome, which is caused by a mutation or deletion in the gene responsible for activity-dependent neuroprotective protein, a molecule that plays a crucial role in brain formation and the development of other organs; Might’s son, Bertrand, had another genetic condition called NGLY1 deficiency.

Might had been building an artificial-intelligence-driven system called mediKanren to help find treatments for rare diseases, including Bertrand’s. Part of the Translator project funded by the National Center for Advancing Translational Sciences, the program works to harvest biomedical data from diverse sources, structure it, and connect the dots. It’s looking for what Might calls “unknown knowns”: drug interactions that have already been revealed by science but remain buried deep in millions of data points.

“It looks like low-dose ketamine is an upregulator for ADNP,” wrote Might. “Do you think this makes sense for ADNP patients?”

On his phone, Davis did a quick literature search. “It was an absolute lightbulb moment,” he said. “I was immediately struck by how exciting a potential treatment option this was for ADNP.” In addition to broad disturbances in sleep, vision, cardiac health, and digestion in many patients, the syndrome is characterized by intellectual disability, speech and motor delays, and autism symptoms. Most drugs for neurodevelopmental disorders act at the level of the synapse by changing neurotransmitters, Davis explained. But Might’s system suggested that ketamine could alter ADNP gene expression — addressing the syndrome almost at its root.  

“In these patients, that gene is sort of running at 50% capacity: One copy’s broken, the other copy’s working,” explained Might. “So the tool, when it sees that, says, ‘OK, well, maybe I can make the functioning copy work twice as hard to compensate.’”

Davis soon found studies that seemed to confirm the premise. A 2017 paper pulled up by the mediKanren platform found that low doses of ketamine induced the expression of ADNP in mouse cancer cells. Another 2015 paper showed the same relationship in rat neurons.

“This is an incredibly rare disease. There just weren’t many people looking outside the box in terms of ways to treat it,” said Davis. “So I don’t think it came on anyone’s radar because the number of people looking was so small.”

Davis, also the chief scientific officer for the ADNP Kids Research Foundation, shared what he had with its founder, Sandra Sermone. In 2014, her son Tony was the first person to be diagnosed with ADNP syndrome in the United States, shortly after a group of researchers in Belgium first described the syndrome.

“It seemed too good to be true,” said Sermone. “I took it with a grain of salt, and we just started investigating everything.”

The foundation had been down this road before, with a promising peptide drug called NAP discovered by researchers at Tel Aviv University alongside ADNP in 1997. The foundation tried to kick off a study in ADNP syndrome with Coronis NeuroSciences, an Israeli company formed to commercialize the peptide, but the planned trial hadn’t materialized.

Ketamine had the potential to be different, though. Unlike a novel peptide that would need to be formulated and packaged, generic ketamine is cheap to access and already in wide use as an anesthetic. There has also been a resurgence in research on the drug for treatment-resistant depression and other mental health conditions. In 2019, Johnson & Johnson received FDA approval for a form of ketamine delivered by nasal spray.

The drug has risks, including dissociative effects that can impact patients differently. Subanesthetic levels of the drug may mildly increase blood pressure, and higher doses used for anesthesia have been associated with cognitive deficits when used in very young children. But the drug’s fairly well-understood safety profile made ketamine a lead worth following.

So in April 2019, Sermone brought their research to Joseph Buxbaum, head of the Seaver Autism Center for Research and Treatment at Mount Sinai in New York. Buxbaum and his colleagues dug into the papers on their own — and soon, Alex Kolevzon, a child and adolescent psychiatrist and clinical director at the center, was designing a trial and working to get it approved. By May, the center was ready to start recruitment.

While Sermone was working to kick off the trial, she thought her son would be eligible; he was just young enough to make the age cutoff. But in March, as the pandemic’s first wave was building, Tony developed a fever that gave way to seizures, then failing lungs and kidneys. He had to be put on a ventilator. He survived after nearly two weeks in the hospital, but it became too risky for him to take an experimental drug.

“It’s a blessing and a curse,” said Sermone. “If my son was in it, because I’m one of the ones who found it, people would think that I got preferential treatment.”

In the end, the trial recruited 10 kids ages 5 to 12. It was another child just at the very top of the age threshold who filled in one of the slots: Malvagno’s son Mateo.

“I was on the fence about doing this trial to begin with,” said Malvagno. Like the other parents of trial participants, who have the right to consent on behalf of their children, she had to weigh the potential benefits against the downsides of uprooting Mateo’s care and routines.

For 40 minutes, each of the kids would have to sit with an IV in their arm as the drug pumped into their veins, and return for three rounds of follow-up tests over the next month. The parents would fill out standardized assessments of symptoms like irritability, hyperactivity, and social withdrawal. Clinicians would mirror those with their own list of measures. “We wanted to get a battery of more objective assessments,” said Kolevzon, including electrophysiological and eye tracking studies. And critically, they’d test blood to let them see whether ADNP expression actually increased. 

“Who wants to put their child through weekly blood draws or a 40-minute infusion?” said Malvagno. 

None of those visits would be easy to manage. “My kid is like a bull in a china shop,” said Malvagno. “He has to be anesthetized for people to look at his teeth, and when he comes out of it, he goes staggering down the hospital hallways. No way we’re going to get him to sit for these tests with the little EEG cap.”

To prepare the kids and their parents, the Seaver center sent stickers to practice applying to the kids’ chests, so they could get comfortable with the sensory experience of ECG readings. Some children practiced wearing a blood pressure cuff ahead of time. During the infusions, the center provided a therapist who played guitar for the kids who are calmed by music.

“I had my doubts about getting through all of it,” said Malvagno, “but we did it, we made it through.” 

In July, the early results came in. The work still has yet to be submitted to or published in a peer-reviewed journal, and the RNA analysis that will show whether ketamine increased ADNP expression is still pending. But there were no reports of serious adverse events, and the preliminary analysis of the parent and clinicians’ assessments showed positive changes one week after the infusion. 

“It was different for different kids,” said Kolevzon, “but we heard improvement around things like hyperactivity, we heard improvement around things like sensory sensitivities. We heard improvement around things like language — not necessarily verbal output, but communication. Those are all extremely important domains that are clinically meaningful.”

As Kolevzon and his colleagues caution, though, “it’s not uncommon for early-stage pilot studies like this to be positive.” That’s due, in part, to the bias brought into play with measures reported by providers and parents. “Any time that your child has as many issues as kids with ADNP syndrome have, and then they’re given a drug, parents become very optimistic about whether it’s helping them,” said Alycia Halladay, chief science officer at the Autism Science Foundation. 

The early results also only analyzed the drug’s impact after one week; the durability of its effect could have an impact on its utility. And while more data may show the drug increases ADNP expression, it’s unclear how big of an impact that would have on brains that developed without enough of the protein. A more direct fix would involve correcting the mutation itself through gene therapy.

For the parents involved, though, it’s hard to temper enthusiasm after the preliminary results. “The hairs on my arms stand up every time I think about it, even now,” said Sermone. 

Several parents told STAT they saw significant changes in their children’s communication and behavior that lasted for four to 11 days after the infusion. “It felt like we weren’t covering up or trying to mask symptoms,” said Amanda Eichenberger, mom to trial participant Zella, who started having success with potty training for the first time in the week after the infusion. “It felt like we were doing something of significance, like a cognitive change that was going on beneath the surface.” 

Mateo received his infusion on a Thursday, and Malvagno saw an immediate change. 

“He was still himself, he was still my mischievous little Mateo,” she said. “But he was calm, and focused.” He stopped his persistent vocalizations, which sometimes last for hours at a time. “I was able to be a parent for the first time, just a parent, not a damage control ninja warrior expert ready to leap tall buildings in order to stop my child from injuring himself.”

On the Sunday after the infusion, they went for a walk on the beach at Coney Island, and at one point, Malvagno gestured for Mateo to come sit with her in the sand — something they never do. “He came over next to me and he sat on the sand, and we lay down for 30 to 40 minutes,” said Malvagno. “It was very emotional for me.” They were both able to simply stop and be together.

Experts emphasize that the results of a small safety study should not be touted as evidence ketamine works. What the trial does signal is a path forward — if not an unobstructed one — for future ADNP syndrome research.

“It was simultaneously a big win for AI — to show that this is possible, and you can have AI come up with a meaningful suggestion — and you can turn it over to a really engaged patient community and they’re willing to run with it from there,” said Might. 

But families affected by rare disease can only run so far on their own. AI got them off the blocks by uncovering the potential of a repurposed drug, and they’ve sprinted through the first clinical trial in record time. To determine whether ketamine is safe and effective enough to use as a treatment for ADNP syndrome, they’ll need to start training for a marathon.

Kolevzon plans to finish up the study’s analysis and submit for publication by the end of the year, and is already planning a randomized controlled trial with blinding.

With more participants, it likely won’t be practical to ask kids and their families to travel to New York for weeks on end. So Kolevzon is considering a different dosing schedule and other delivery modes — including a nasal spray, oral, and subcutaneous delivery — to make a more distributed trial possible. 

“We want to design it in a way that it can actually get approved,” said Kolevzon. “Because at the end of the day, if we’re going to make this available to the families, it needs to be covered by insurance. And in order to get it covered by insurance, it needs to be approved by the FDA.”

The first ketamine trial was unique in its speed and low cost. A virtual charity run organized by the foundation raised $150,000 and paid for the first study in full, with average donations in the range of $20 to $100. “There’s no way we can keep up,” said Sermone, who guesses the next study will cost more than $1 million. The foundation had another fundraiser at the end of July, but these families can only tap their networks so often; this time, they raised only $100,000.

“We need to have Richard Branson or somebody become a sugar daddy,” Malvagno quipped.

More likely will be traditional funding sources, like grants from the National Institutes of Health. If ketamine is shown to upregulate ADNP in the study, it could also catch the attention of scientists and funding organizations invested in research on other disorders. ADNP mutations are one of the most frequent single-gene causes of autism, and the protein regulates more than 400 other genes. 

“That’s always been our strategy,” said Kolevzon. “When you take relatively common forms of autism spectrum disorder, the idea is we’re trying to use those specific genetic forms to understand autism more broadly.” 

Meanwhile, improvements in tools like mediKanren could make this kind of accelerated research pathway more common. Right now, requests to use the system go through Might’s precision medicine institute at UAB. The ketamine lead only ended up in the right hands by happenstance, said Davis. “I happened to know Matt Might, he happened to run ADNP through his system, and all I did was review the literature enough to be able to say, holy crap, this actually is an incredibly exciting potential treatment.”

At the end of July, though, the NIH released a funding call to develop a more public-facing user interface for the translator tools, including mediKanren. Might plans to put in a proposal to build a more usable front end for physician-scientists who want to access the biomedical literature in this structured way. 

Broadly disseminated AI tools may be able to speed up the time it takes to discover a drug worth testing in a clinical trial, and repurposing an existing drug accelerates the process even more. But as the ADNP parents have discovered, even a rapid influx of AI-generated ideas will still leave them playing the waiting game.

“Some of us have waited so long,” said Malvagno, “and I don’t know if Mateo’s going to be available to be in the next trial,” now that he’s aged out of the original inclusion criteria. 

Until they have more answers, Malvagno and other parents remain stuck trying to make treatment decisions based on limited information. “When you have so little evidence-based medicine when it comes to this syndrome, you really start grasping at straws and looking for anything,” said Eichenberger. 

There is now an option that they might not have considered before: off-label ketamine. “Everybody’s thinking about it,” said Kolevzon. “It’s an obvious question.” Though ketamine is nowhere near an FDA approval for ADNP syndrome, doctors can choose to prescribe it for uses other than its approved indications. 

“That always makes it a really, really tough decision because of the ethical considerations,” said Steve Levine, co-founder of ketamine-assisted therapy clinic Heading Health. “Is there sufficient evidence to support this? Is it really possible to fully consent a patient given the desperation, as well as maybe not enough information to really make a good decision?” 

To Sermone and Kolevzon, the answer is clear: While the results are promising, they’re far from a green light. “I advise strongly against trying to pursue ketamine, commercially or clinically, because of safety concerns and the risk of losing our opportunity to study it rigorously,” said Kolevzon. Despite all the progress made in the last two and half years, the ADNP syndrome community still doesn’t have a treatment in hand. 

For the parents whose children participated in the trial, it’s especially wrenching to return to uncertainty. “That full week after the ketamine was a life-changing experience for me as a parent,” said Malvagno. “I want that back.”