Scientists are now using artificial intelligence to discover new medicines at a speed previously thought impossible. This technology is creating potential treatments for conditions that have stumped researchers for decades, including Parkinson's disease, rare illnesses, and infections caused by antibiotic-resistant bacteria.
By rapidly analyzing vast amounts of chemical data, AI is identifying and even designing novel compounds that could change how we fight some of the world's most challenging health problems. This new approach promises to accelerate a process that has traditionally been slow, expensive, and filled with failure.
Key Takeaways
- Artificial intelligence is being used to discover new antibiotics capable of fighting drug-resistant superbugs like MRSA.
- Researchers have identified five promising new compounds that could slow the progression of Parkinson's disease using machine learning.
- AI models are also finding new uses for thousands of existing, approved drugs, offering hope for rare diseases that are often overlooked.
- The technology can screen billions of potential drug molecules in days, a task that would take months and millions of dollars with conventional methods.
A New Front in the War on Superbugs
For years, the effectiveness of antibiotics has been declining as bacteria evolve to resist them. This growing crisis of antimicrobial resistance is responsible for more than 1.1 million deaths annually, a figure projected to exceed eight million by 2050 if new solutions are not found.
Developing new antibiotics has been a major challenge, with only 12 approved between 2017 and 2022. But researchers are now using AI to close this gap. James Collins, a professor at the Massachusetts Institute of Technology, is leading one such effort.
His team has developed AI models that can identify potential antibiotics from massive chemical databases. "We can – in a matter of days or hours – look at massive libraries of chemical compounds to identify those that display antibacterial activity," Collins explained.
By the Numbers: AI vs. Traditional Methods
- Speed: AI can screen billions of molecules in a few days.
- Traditional Speed: Screening one million molecules typically takes six months.
- Cost: AI screening costs a few thousand pounds.
- Traditional Cost: Conventional screening costs several million pounds.
The team trained a generative AI to understand the chemical structures of known antibiotics. They then tasked it with screening over 45 million compounds to find ones effective against highly resistant bacteria, including those that cause gonorrhoea and MRSA.
The AI didn't just find existing chemicals; it designed 36 million entirely new compounds. From this massive pool, the team synthesized 24 in the lab. Two of them proved highly effective at killing strains of bacteria that are resistant to current drugs. These new compounds appear to attack the bacteria in novel ways, which could make them a powerful new class of medicine.
Targeting Parkinson's Disease at its Source
Parkinson's disease was first identified over 200 years ago, yet there is still no treatment that can slow or stop its progression. The disease affects more than 10 million people worldwide, and its cause remains a subject of debate among scientists, making it difficult to develop effective drugs.
Michele Vendruscolo, a professor at the University of Cambridge, is using machine learning to tackle this problem from a new angle. His work focuses on the clumps of misfolded proteins, known as Lewy bodies, that build up in the brains of Parkinson's patients and are thought to contribute to nerve cell death.
"We can analyze this data and make very accurate predictions about the way candidate molecules will bind to the target at a scale that was unthinkable until a few years ago."
Michele Vendruscolo, University of Cambridge
Vendruscolo's team used an AI to search for small molecules capable of targeting these protein clumps. The AI was trained on a set of known compounds and then used that knowledge to propose new, potentially more effective ones. The system rapidly narrowed a search that would otherwise be impossibly large.
"The number of possible small molecules is far larger than the number of atoms in the Universe," Vendruscolo noted. After the AI made its predictions, the team tested the suggested compounds in the lab and fed the results back into the system, allowing it to learn from its mistakes. This iterative process led to the identification of five promising new compounds that are now undergoing further testing.
The ultimate goal is even more ambitious. Vendruscolo hopes to use AI to find molecules that can prevent the proteins from misfolding in the first place, effectively stopping Parkinson's before it starts.
Unlocking New Potential in Old Drugs
Creating a new drug from scratch is a long and expensive journey. An alternative approach gaining traction is drug repurposing—finding new uses for medicines that have already been approved for other conditions. This strategy significantly reduces costs and timelines because the safety of these drugs is already established.
The Story of David Fajgenbaum
David Fajgenbaum, a professor at the University of Pennsylvania, saved his own life by repurposing a drug. Diagnosed with a rare disorder called Castleman disease, he was running out of options. After extensive research on his own blood samples, he identified sirolimus, a drug typically used to prevent organ rejection in kidney transplant patients, as a potential treatment. It worked, and his disease has been in remission for over a decade. This experience inspired him to found Every Cure, a nonprofit that uses AI to find new uses for existing drugs for thousands of other diseases.
AI is a powerful tool for this kind of work. At Harvard Medical School, an AI model analyzed existing data and identified nearly 8,000 approved drugs that could potentially be repurposed to treat 17,000 different diseases.
This method is particularly valuable for rare diseases, which often receive little funding from pharmaceutical companies. Researchers at McGill University used AI to find treatments for Idiopathic Pulmonary Fibrosis (IPF), a progressive lung disease. Their AI created a "virtual disease system" to simulate how the disease advances in lung cells.
"If we can figure out how the cell went from healthy to abnormal, maybe we can reverse it, or slow it down," said Jun Ding, an assistant professor who worked on the project. By testing different drugs within this virtual model, the team identified eight potential treatments, including a common hypertension drug that could be a low-cost option for IPF patients.
The Future of Medicine is Driven by Data
Companies like Insilico Medicine are already bringing AI-discovered drugs to clinical trials. Their candidate for IPF, Rentosertib, was designed entirely by AI and has shown positive results in early trials. If successful, it could be available to patients by the end of the decade.
However, challenges remain. Much of the valuable data on drug properties is owned by private companies and not publicly available. AI is currently best at the initial stages of discovery, and the full process of developing, testing, and approving a new medicine is still long and complex.
Despite these limitations, the impact is already being felt. As Jun Ding stated, "My belief is, in the next five to 10 years, the majority of new drug development could be guided by AI, or even entirely based on AI." The technology is not just speeding up old methods; it is creating a new paradigm for medical discovery, offering hope for millions of patients worldwide.