A year ago, I heard that Joseph Coates had only one thing to decide. Did he want to die at home or hospital?
Living in Coates (37 years old) and Wash. Renton was rarely conscious. For several months, he was fighting a rare blood disorder of poetry syndrome, which led to a height of paralysis, feet and heart, and failed to kidney. Every few days, doctors had to drain the liter from the abdomen. He is one of the only treatments that can alleviate him because he is so painful to receive stem cell transplantation.
“I gave up,” he said. “I just thought the end was inevitable.”
But Coates’ girlfriend, Tara Theobald was not ready to quit. So she sent an email to ask a doctor called David Fajgenbaum in Philadelphia.
The next morning, Dr. FAJGENBAUM replied and proposed a non -traditional combination of chemotherapy, immunotherapy and steroids that were not previously tested with the treatment of coates’ disorders.
In a week, Coates responded to treatment. In four months he was healthy enough for stem cell transplantation. Today he is mitigating.
There was no doctor or anyone who saved life. I spit it by an AI model.
In laboratories around the world, scientists use AI to search existing medicines for effective treatment for rare diseases. Although the use of drugs is not new, the use of machine learning is speeding up the process, and it can expand the possibility of treating people with rare diseases and options.
Thanks to the version of the technology developed by Dr. FAJGENBAUM’s team of Fajgenbaum from the University of Pennsylvania, the drug is rapidly changing its use of rare and aggressive cancers, fatal inflammatory disorders and complex neurological conditions. And often they are working.
So far, a small number of success stories have led the researchers to ask questions.
“There is a drug of treasure that can be used for many other diseases. We have no systematic way to look at it. You can already buy it at the pharmacy. “
The National Institute of Health defines rare diseases as affecting less than 200,000 people in the United States. But there are thousands of rare diseases, which affects at all. Tens of millions of Americans and Hundreds of people World.
But more than 90 % Rare There is no approved treatment, and the pharmaceutical giant does not make a lot of resources to find it. Christine Colvis, who leads the NCATS drug development partnership program, said that there is not much money in developing new drugs for a small number of patients.
Dr. Marinka Zitnik, an associate professor of Harvard Medical School, who studies computer science applications in medical research, said that this is the reason for making drug riposis that uses a “seductive alternative” path to find the treatment of rare diseases. Dr. Zitnik’s Harvard Lab has built another AI model for drug use.
Dr. LO said, “Other laboratory discovery technologies have already been put in the map.” AI just puts a rocket booster. “
Find clues in old studies
Use is quite common in constraints. MINOXIDIL, developed as a blood pressure drug, has been used to treat hair loss. The Viagra, which is originally sold to treat the heart, is now used as an erectile dysfunction. Semaglutide, a diabetic drug, is best known as the ability for people to lose weight.
When Dr. FAJGENBAUM first conveyed the use of the drug, it tried to save his life. While he was at the age of 25, he was diagnosed with a rare sub -type of disorder called CASTLEMAN disease, which resulted in an immune system that landed him in the intensive care room.
There is no other way Treat the Castleman diseaseAnd some people do not respond to the available treatments. Dr. Fajgenbaum was one of them. During the temporary inpatient and chemotherapy, Dr. FAJGENBAUM attempted non -traditional treatment by conducting tests on his blood, dealing with medical literature, and treating them.
“I have received this clear realization of $ 1 billion and 10 years to make new drugs from the beginning,” he said.
Dr. FAJGENBAUM’s life is a common drug called Sirolimus, which is generally provided to kidney donors to prevent rejection. The drug has relieved his Castleman disease for more than 10 years.
Dr. FAJGENBAUM continued to be a professor at the University of Pennsylvania and began to find other drugs for unknown purposes. He said existing research is full of overlooked clues about the potential relevance between drugs and treated diseases. “If they are in the published literature, someone should not find this every day?”
His laboratory has been a bit of success, discovering that a new cancer drug has helped other Castleman patients. But the process was hard, and his team demanded to investigate “a drug and a disease at a time.” Dr. FAJGENBAUM decided to speed up the project. In 2022, he established a non -profit organization called Every Cure, which aims to compare thousands of drugs and diseases at once using machine learning.
Works similar to all treatments occur in other laboratories around the world. In Penn State Stanford University and Japan and China.
In Alabama, Burmingham, the AI model suggested chronic vomiting through an iso profile alcoholic nose and proposed a deceased 19 -year -old patient. Matt Mill, a professor of University of Alabama at the University of Alabama, has shown all the treatments that have been in the history of drugs for nausea, “In essence,” leading the research institute that developed this model.
Alcohol said, “It popped up at the top of our list.”
Dr. The model developed by MOWN’s Institute has successfully predicted other treatments. In general, amphetamine, which is used to treat ADHD relaxation periodic paralysis, has rare genetic disorders. Parkinson’s drugs helped patients with neurological conditions moved and talked. The general blood pressure medicine called Guanfacine greatly improved the mobility of pediatric patients with other neurological conditions.
Many drugs are doing more than two things, May said. Their additional features are sometimes characterized by side effects. “If you comb the enough drugs, you will eventually find the side effects you are looking for,” he said.
Dr. FAJGENBAUM’s platform at the University of Pennsylvania compares about 4,000 drugs with 18,500 diseases. For each disease, pharmaceuticals obtain scores according to the possibility of efficacy. If the prediction is made, a researcher will find a promising idea through them, then perform a laboratory test or connect with the intention to test the drug.
In other places, pharmaceutical companies are completely discovering new drugs using AI. Companies are already billions of dollars. But drug use will not be found to be favorable to a party. Many drug patents expire decades, and AIDEN HOLLIS, a professor of economics at Calgary University, focuses on medical commerce, said that there are few incentives for pharmaceutical companies.
Once the drug becomes one Thousands of generics It has been approved by the Food and Drug Administration, and is generally lowering the price in the face of intense competition.
Dr. Fajgenbaum said, “If you use AI to prepare a new drug, you can make a lot of money with new drugs.
In order to support venture funds, all treatments were promised more than $ 100 million last year. Ted’s bold project and Advanced Research Project Health AgencyIn order to support potential research innovation, the agency within the Federal Health Ministry. Dr. FAJGENBAUM said that all treatments will partially use this money to fund the clinical trials of drugs with use.
Dr. Grant Mitchell, a co -founder of Dr. FAJGENBAUM, and a classmate in medical school, said, “This is an example of AI that we don’t have to be afraid, and we can really be excited.” This will help many people. “
‘Someone had to try first’
Dr. Luke Chen was skeptical when he suggested that Dr. FAJGENBAUM’s models use ADALIMUMAB, a drug commonly used to treat arthritis, cron disease and ulcerative colitis.
“I didn’t think it was effective.”
However, the patient has already suffered chemotherapy and bone marrow transplantation and attempted drugs, including drugs that saved Dr. Fajgenbaum’s life. Nothing worked and he was in the hospice.
Dr. Chen said, “We gave up basically, but we made the last call to David.
Dr. Chen, who has no other options, gave the patient Adalimum. In a few weeks, the patient was mitigating. This case was a recent topic Paper New England Journal.
Dr. Zitnik said the model could not collapse. AI sometimes can predict “based on not enough evidence.”
Dr. Colvis said that the potential treatment ranking due to success can also be difficult. Such a problem makes the doctor’s supervision important. Sometimes doctors will determine that the treatment suggestion is too dangerous, she said. Dr. Colvis said, “But they see something and say,” Okay, this seems reasonable. “
Dr. FAJGENBAUM first suggested that Dr. GAO is suspected when Dr. Wayne Gao, a Washington blood scholar and tumor specialist, attempted a new treatment for one of his patients.
The patient was Coates, the Washington man headed to the hospice, and the aggressive drug combination proposed by Dr. FAJGENBAUM’s model was “a little crazy.” In fact, he was worried that the treatment would kill Koate faster.
But Coates were young and there was no other treatment to consider. So Dr. Gao said, “Someone had to try first.”
A year after the brush last month, Coates and his girlfriend thanked me for visiting Dr. Fajgenbaum in Philadelphia. The smiling coated was a picture of health. He lasted his muscles since he met a doctor.
Coates adjusted the ankle during the morning exercise that morning. But if not, he said, “It’s okay.”
Maxwell Straban Contribution.
