Two pharmaceutical companies have embarked on a brave new world, having begun human testing for the first time on a drug treatment for obsessive-compulsive disorder designed by artificial intelligence.
British startup Exscientia and Japan’s Sumitomo Dainippon Pharma used artificial intelligence to create the drug in less than 12 months, cutting four years from the average time it takes ordinary humans to develop a medication.
Exscientia CEO Andrew Hopkins described the clinical human trial of the drug — a molecule called DSP-1181 — as a “key milestone in drug discovery.”
“Our driving motivation is to accelerate the range of innovative drugs from cutting-edge science entering into the clinic to increase the treatment options for patients. That means reducing the time to make and test a drug. The consumer should see benefits from faster progress to the clinic,” Mr. Hopkins told The Washington Times.
Artificial intelligence has been used to quickly and accurately diagnose diseases and analyze patient data, but this time it had a more robotic “hands-on” role in the creation of a medicine.
The AI created the drug by using algorithms that sifted and sorted through compounds to determine the safest and most effective one for treating a specific disease.
“Different properties of molecules will create different side effects, and so it’s a very intriguing idea to use AI, a computer, to help predict,” said Dr. Wendell Gibby, a radiologist in Utah and CEO of Novarad, a health care technology and imaging company. “AI is a very powerful technique that’s being used across many fields of medicine, especially in radiology.”
Exscientia used a machine-learning platform called Centaur Chemist, which reportedly trims off years needed to research new compounds by pairing AI methods with knowledge about how medicines interact with the human body, according to news website Tech Acrobat.
The tool can study millions of molecular combinations — a much quicker alternative to human scientists who “operate in the real world,” the tech news company wrote.
Because of the artificial intelligence, the candidate compound for the drug to treat obsessive-compulsive disorder was found in a search of 350 synthesized compounds rather than the average 2,500 compounds.
DSP-1181 entered phase 1 trials in Japan at the end of January. If the trials are successful, plans to develop the drug globally, including for the U.S., will be underway.
Making a single drug that reaches phase 1 testing approval can cost millions of dollars, but developing candidate drugs faster with AI can lead to significant cost savings.
Companies have used artificial intelligence for an array of health care services, including detection of breast cancer, eye diseases and acute kidney injuries. It also can complete administrative tasks for physicians.
Daniel Faggella, founder and CEO of Emerj, an AI research company, said artificial intelligence could help extend specialized health expertise in diagnostics and other areas.
“Whether [AI] be good, bad or ugly, it’s going to be slow in health care,” Mr. Faggella said. “On the aggregate, maybe that slowness will help because nothing’s going to hit the scene too quickly and disrupt things too quickly.”
He said he expects to see gradual improvements across the board in health care — including with diagnoses, genetics and efficiency in treating patients — thanks to AI.
As artificial intelligence plays a more active role in health care, some question the way this technology operates and its impact on the human labor force.
Dr. Jennifer Joe, an urgent care physician and founder of MedTech Boston, expressed concerns about the algorithms AI uses and the data available.
“Our biggest concern is bias in data,” said Dr. Joe. “I think we’ve seen many examples of that. A lot of the data we’ve collected, especially from clinical trials, has been from men, traditionally white men, so how much of that data are you going to be able to extrapolate out to women or people of color or to pregnant women or to older people?
“We want to make sure the algorithm is fair, is unbiased, is regularly vetted and checked against best practices using clinical data and with clinicians,” she said.
Although AI can outperform humans in many tasks, researchers concluded that jobs won’t be automated on a large scale for a while, according to an article published last year in the Future Healthcare Journal.
The researchers, who hail from Babson College and Deloitte Consulting, pointed out ethical implications around the use of AI and noted that many AI systems are built for single tasks whereas health care professionals do more than read and interpret images.
Mr. Hopkins of Exscientia acknowledged that AI has its limitations. He said it cannot predict all the possible ways a drug will act in an “incredibly complex” human body, but it can help researchers learn faster and adapt their experiments.
“If we can learn faster from fewer experiments, we can move drugs forward to the patient faster than we have before,” he said. “Drug discovery is about embracing uncertainty, and AI is helping us learn faster in the face of that uncertainty.”
• Shen Wu Tan can be reached at stan@washingtontimes.com.
Please read our comment policy before commenting.