By Laura Freeman

“Since mapping the human genome, we’ve identified 10,000 and counting genetic factors influencing diseases. In some of those conditions, particularly where a single gene controls changes, we are making real progress in using what we’ve learned to improve outcomes,” Matt Might, PhD, Director of the Hugh Kaul Precision Medicine Institute at the UAB Heersink School of Medicine, said.

One case in point is an already approved drug now showing promise in treating symptoms in a form of autism related to a rare genetic mutation that causes a deficiency in activity-dependent neuroprotective protein (ADNP).

Matt Davis, MD, a neurosurgical resident at UAB, asked the precision medical team for help with his son, Benjamin, who had been born with ADNP syndrome. Because children with this syndrome have only one good copy of this gene, they don’t produce enough of this protein. According to the NIH, ADNP deficiency causes developmental delays and other symptoms found in autism. It can affect fine motor skills, hearing, vision, muscle tone, the heart, the immune, endocrine and gastrointestinal system, as well as the brain and behavior.

“We wanted to help Benjamin, and this seemed to be the perfect challenge for mediKanren, the AI tool our precision medicine institute developed with the help of funding from the National Center for Advancing Translational Science,” Might said.

In what may be the first time an AI tool produced enough evidence to gain approval for an FDA clinical trial, the precision medicine team began loading all data known about Benjamin’s case and that of a second patient, then instructed it to explore all available medical literature to seek out therapeutic agents that might be beneficial in treating the disorder.

The AI returned a strong response—low dose ketamine is an up-regulator of the protein governed by the ADNP. It has been used as an FDA approved safe and effective therapy for decades.

“Ketamine was originally known for its use in veterinary medicine anesthesia and is now being used in humans for pain management and depression,” Might said.

After identifying low dose Ketamine as a potential therapeutic agent, the Precision Medicine Institute sent its findings to the ADNP Kids Research Foundation, which relayed it to the Seaver Autism Center For Research And Treatment at Mount Sinai Hospital. The center ran a preliminary clinical study that confirmed that ketamine is effective in increasing protein level. The response was strong enough that a broader clinical trial is in development.

Another area where the crossroads of genetics and AI could make a difference in the course of custom-tailored care for individual patients is cancer treatment.

“Each tumor has its own unique genetic code, and the code may be different in a second tumor in the same patient,” Might said. “It may also continue to mutate over time and become a moving target. If we can sequence the unique code of a tumor, we can use our AI tool to look for the treatment strategies most likely to be effective against it. Then we can sequence it again as needed to monitor changes.

“When you see the genetic sequence for a cancer cell, you may be surprised. For example, we had a patient with an aggressive prostate cancer. We expected it to come back identified as prostate cancer, since that was where it was located. But this particular sample had a profile that looked like a cancer usually found only in women. So we adjusted therapy to use the same strategy we would use in a female patient. It worked.”

The list of genetic diseases humans are heirs to is long, and whether a possibility becomes a probability sometimes depends on changes brought on at the molecular level by epigenetic factors and other influences that can turn a blueprint for trouble into an active attacker. Even in disorders where a genetic causal relationship hasn’t been established, how each patient responds to a treatment can vary according to the specific genetic profile. Beyond helping clinicians choose the approach to treatment most likely to be successful, an AI analysis of a patient’s genetic profile can also provide valuable clues in undiagnosed illnesses.

“Some diseases are difficult to diagnose and we see quite a few cases that are unlike anything we’ve ever seen before,” Might said. “AI can give us clues about what might work and parallels to similar conditions and what worked for them. Most of all, it tells us not just what the genetic sequence is, but what it might mean.”

From a background in computer science, Might was led to a career in precision medicine by a personal quest to put a name on his own son’s illness and find a treatment for it. Bertrand was the first patient in the world identified has having an NGLY1 deficiency. After working in the White House in President Obama’s Precision Medicine Initiative, Might came to UAB to lead the institute. Now he and his team are helping others find answers at the end of their quest and perhaps discover a pathway to better outcomes. Better yet, precision medicine is now identifying more genetically-linked disorders early enough to change the course of their progression and prevent a genetic predisposition from turning into a disaster.

“My genes make it more likely that I could develop macular degeneration, so I’m doing everything I can to keep my eyes healthy and monitor for early signs of changes,” Might said. “Other people may find that they are predisposed to heart disease, diabetes, a specific type of cancer, dementia or a whole list of other disorders. Knowledge is power, and learning the risks early gives us an opportunity to rewrite our future. That is the ultimate promise of precision medicine.