Time might pass quickly as you pace your drugstore, waiting for someone behind the counter to prepare your prescription. But have you ever considered how long it takes to develop a new drug?
A medicine takes around 12 years and $1 billion to develop. Artificial intelligence (AI) attempts to make this happen more quickly.
The objective is to increase the number of medications available for uncommon illnesses. There are fewer side effects. Treatments that are unique to you. In less time and for less money.
At least 230 businesses are using AI to assist in the development of new medicines. Despite the hoopla, artificial intelligence has failed to bring medicine to market on its own. Will it be able to keep its word and save lives?
New therapies are desperately needed by patients like Laura Roix.
Time is running out
Roix had a cough every October. Her doctor treated her for pneumonia every year. However, the cough would constantly come back. She went to a pulmonologist, a specialist who specializes in the lungs because she had a family history of lung illness.
Scars were seen at the bottom of her right lung during a scan. There was nothing to be concerned about; it was most likely asthma, she was informed. Roix depended on inhalers and lozenges for the next six years to relieve her dry, hacking cough.
It didn't get any better, though. In fact, she began to run out of breath.
Something was awry, Roix sensed. As a result, she went to another pulmonologist. Finally, she received the correct diagnosis: idiopathic pulmonary fibrosis, a life-threatening lung disease (IPF). IPF scars and stiffens the lungs over time, making it difficult to breathe. If the condition worsens, a lung transplant may be required.
Roix used oxygen at night for the following few years. She enrolled in pulmonary rehabilitation. Her cough had worsened to the point that she couldn't even converse on the phone.
"It's a never-ending loop," she explains. "You are unable to take a breath. You're sneezing and coughing. You're coughing a lot more now. You get anxious as a result. And it's all a jumbled mess."
There were no viable alternatives. "There was nothing on the market when I was diagnosed with idiopathic pulmonary fibrosis," Roix adds.
She enrolled in a new medication clinical study and was placed on the lung transplant waiting list. Roix even traveled to Washington, DC, to lobby the FDA to approve two IPF medicines that are currently available: nintedanib and pirfenidone. They weren't, however, her solution.
Taking on a Life-Threatening Lung Disease
Roix is unknown to Alex Zhavoronkov, PhD. However, the CEO of the biotechnology company Insilico Medicine does not want anybody else to go through what she has.
Zhavoronkov disclosed a possible cure for IPF on his 42nd birthday, which was created by Chemistry42, his company's AI platform.
The news from Insilico is one of the most recent in the field of AI and drug development. However, the most difficult problems in treating persons like Roix remain.
Improved therapy is one of AI's promises. Insilico claims that its new medication, IPF, requires a tenth of the dose of an existing treatment, nintedanib, and acts in a different way.
Because of fibrosis, the mechanism through which your body creates scars, the business picked IPF as the condition to aim to treat using AI.
"Fibrosis is really beneficial to your body. Your scars and wounds will not heal if you do not do so "According to Zhavoronkov. Fibrosis, on the other hand, goes awry in IPF. Our, he believes, is one way that bodies degrade over time. IPF is increasingly common as people become older, although it is not a natural aspect of aging.
In 18 months, Insilico's team created a new medication. First, AI was used to research IPF and discover a new way to combat it. It then used algorithms to discover novel medicines, which it then tested in a lab. It required several rounds of experimentation, with the algorithm learning from each one, as predicted.
As a consequence, a new medication is expected to enter clinical trials in early 2022. If all goes well, it might be on the market in four years. Insilico has yet to publish its IPF findings in a peer-reviewed publication.
The medicine was created with the assistance of artificial intelligence. Clinical studies, on the other hand, are frequently lengthier and more difficult. Many medicines that appear promising in labs and occasionally in mice fail when tried in humans.
AI is like a Ferrari, as Zhavoronkov puts it: "You swiftly go from 0 to 100. However, you must then travel at the speed of the traffic."
Even if AI doesn't succeed in perfecting every new medication, it can make the process easier.
"At the very least, you can automate and make much more systematic some of the jobs that now rely on human intuition," says Regina Barzilay, PhD, a Massachusetts Institute of Technology professor of AI and health.
She was part of a team that employed artificial intelligence to develop a new antibiotic. Halicin was named after HAL, the computer from 2001: A Space Odyssey. The researchers stated in the journal Cell that "the moment is ripe" for this research.
There are drawbacks to AI, despite its potential. A balance must be struck between data, human biology, and how far technology may go before human intervention is required.
How to Pick a Lock
Consider a lock for which you require a key. That's how Ola Kalisz, a research engineer at ExScientia in the United Kingdom, describes the process of looking for novel pharmaceuticals.
The lock is a symptom of an illness. Treatment is the key. "You try to reverse-engineer what would be the key that cracks open the lock after being handed the lock," she explains.
To do so, AI systems learn by trial and error, selecting fresh data points from which to learn, predicting which existing medicines might work, or recommending how to create a new one.
"You simply sit at your computer and code away at the concept that's in your brain," Kalisz explains. Virtual experiments are used to see if it works. "You want to be sure, 'Did you accomplish your goals?'" Kalisz says.
But first, you'll need a lot of data, the correct sort of data. Alternatively, the road to a novel therapy might lead to a dead end.
Problems with Data
The choices are restricted if there isn't enough data to look through and learn from.
Scientists sometimes can't obtain the data they need because it isn't open to the public. Furthermore, biological data is frequently "standardized," according to Andreas Bender, PhD, of the University of Cambridge's Centre for Molecular Science Informatics. That is, it is insufficiently diversified.
Misses are also possible. "We don't understand the biology of the illness sufficiently to collect the proper data," adds MIT's Barzilay, especially with complicated disorders. There isn't just one lock that requires a key in these circumstances, but a swarm of them.
Bender points out that scientists in this field frequently know more chemistry than biology. Finally, he argues, it's critical to distinguish between utilizing artificial intelligence to develop a chemical and really identifying a cure that humans can use.
When Humans Enter the Picture
The first "AI-discovered" medicine began clinical testing in January 2020, according to reports. It's a medicine for obsessive-compulsive disorder (OCD) developed by ExScientia and Sumitomo Dainippon Pharma, a Japanese pharmaceutical firm. (For this report, ExScientia declined to offer an update on those studies.)
Humans were undoubtedly engaged in the development of OCD and IPF medicines. People are still required to discover and manufacture new medicines. ExScientia's Centaur Chemist platform is called after a legendary creature that is half-horse, half-human, and it combines artificial and human intelligence.
Chemistry42, Insilico's platform, has a name that would make any sci-fi lover happy. Super intelligent mice in Douglas Adams' The Hitchhiker's Guide to the Galaxy build a computer named Deep Thought to figure out the meaning of the cosmos. The answer given by Deep Thought is 42.
The number has become an Easter egg allusion to artificial intelligence for Zhavoronkov and others in the AI area.
When will AI be able to develop medicine on its own? "The challenge is when it can be done really quickly and without people," adds MIT's Barzilay. "And this is something I can't really foresee."
Medication wasn't the answer for Roix in the end. She continued in the clinical experiment until her left lung was transplanted. Her cough has significantly improved. She was even able to run a 5K just 8 months following the surgery.
An organ transplant, on the other hand, is the last option. Many people never get to the top of the waiting list.
In the instance of Roix, her right lung is still in jeopardy. She says, "It's practically dead."
When you're unwell, all you want to know is what will make you feel better. It's unlikely that you're thinking about whether AI created it.
Post a Comment