For centuries, people have chewed willow bark to relieve pain, but it wasn't until the 1800s that scientists from the chemical company Bayer isolated the active ingredient and eventually patented an improved version of it as aspirin.
Aspirin is just one example of a medicine derived from natural sources. In fact, the World Health Organization estimates that about 40% of modern medicines are derived from treatments used by our ancestors.
Despite this incredible success in harnessing nature's bounty, scientists estimate that only a small number of natural compounds have been discovered that could be developed into powerful medicines.
One reason is that identifying, isolating and testing molecules from nature is more complex and time-consuming than synthesizing new compounds in a lab.
Viswa Korlu, an early employee at Recursion Pharmaceuticals, which went public in 2021, decided that AI and other technologies could speed up the process of discovering new medicines from nature.
In 2019, Korrul left Recursion to found Embeda Biosciences, a Boulder, Colorado-based biotech company that chemically analyzes plants to discover potential medicines.
Colleur told TechCrunch that Embeda leveraged all the digital information from around the world about how people in different cultures have used plants to treat pain and illness.
“We found that geographically distant cultures around the world, who have never spoken to each other, are much more likely to use similar plants for similar ailments and symptoms,” he says. “They discover that certain plants are good for stomach aches, certain plants are good for fevers and headaches. This is literally thousands of years of empirical human wisdom.”
Currently, the company's database contains 38,000 medicinal plants associated with approximately 12,000 diseases and conditions.
Once Enveda's AI has identified the plants most likely to have therapeutic effects, it collects the material and tests it using the company's AI models. Unlike traditional methods that study individual molecules, Enveda's transformer models can decipher the “chemical language” of the entire sample.
“Once we know the shape of the molecules, we can prioritize the right set of molecules and say this could one day become a drug,” Korrul said.
Embeda's approach is starting to bear fruit: Two of the company's medicines, one for eczema and the other for inflammatory bowel disease, are scheduled to enter clinical trials later this year, according to Corullu.
The company's scientific breakthroughs have caught the attention of investors. Embeda on Thursday announced it had raised $55 million in a Series B extension from new investors including Microsoft, The Nature Conservancy, Premji Invest and Lingot Investment Fund, and existing investors including Kinnevik, True Ventures, FPV, Rebel Ventures and Jazz Venture Partners. The new funding brings the company's total capital to $230 million.
The extension round allows Embeda to add a long-term strategic partner to its capital table, and the company plans to raise Series C funding later this year after clinical trials begin, according to Corullu.
Microsoft will also provide cloud credits as part of the deal, which are separate from the cash investment, Corll said.
Harvesting plants to find medicines is an old technique, but Embeda is one of the few companies doing it with the help of AI. UK-based Pangea Bio is also studying plants to find treatments for neurological disorders.
Of course, much of the focus in this field is on marijuana, whose natural sources are best known for producing psilocybin, the compound found in so-called “magic mushrooms” and other hallucinogens that may cure mental illness, but Embeda isn't interested in studying those compounds.
“Everyone's focused on cannabis and psychedelics, but these are only a small part of the natural world,” says Corlul. “The natural world is so rich in chemical diversity and biological effects that by studying just a few hundred plants, we could find more potential drugs than we know what to do with.”