Harnessing AI to discover new drugs inspired by nature

Artificial intelligence (AI) is able to recognize the biological activity of natural products in a targeted way, as researchers at ETH Zurich have demonstrated. In addition, AI can find molecules that have the same effect as a natural substance but are easier to manufacture. This opens up huge possibilities for drug discovery, which could also rewrite the rules of pharmaceutical research.

Nature has a vast supply of medicinal substances. "More than 50 percent of all current drugs are inspired by nature," says Gisbert Schneider, professor of computer-aided drug design at ETH Zurich. Nevertheless, he is convinced that we have only exploited a fraction of the potential of natural products. Together with his team, he has successfully demonstrated how artificial intelligence (AI) methods can be used in a targeted way to find new pharmaceutical applications for natural products. In addition, AI methods are able to help find alternatives to these compounds that have the same effect but are much easier and therefore cheaper to manufacture.

Target molecules of natural substances

The ETH researchers are thus paving the way for an important medical breakthrough: we currently only have about 4,000 fundamentally different drugs in total. In contrast, estimates of the number of human proteins are as high as 400,000, each of which could be the target of a drug. There is good reason for Schneider to focus on nature in his search for new pharmaceutical agents. "Most natural products are by definition potential active ingredients that have been selected by evolutionary mechanisms," he explains.

While scientists used to scour collections of natural products in search of new drugs, Schneider and his team have flipped the script: they first look for possible target molecules, usually proteins, of the natural products to identify pharmacologically relevant compounds. "The chances of finding medically meaningful pairs of active ingredient and target protein are much higher with this method than with conventional screening," Schneider explains.

Tested with a bacterial molecule

The ETH chemists tested their concept with marinopyrrole A, a bacterial molecule known for its antibiotic, anti-inflammatory and anti-cancer properties. However, little research has been done to find out which proteins in the human body this natural substance interacts with to produce these effects.

To find possible target proteins for marinopyrrole A, the researchers used an algorithm they developed themselves. Using machine learning models, the algorithm compared the pharmacologically interesting parts of marinopyrrole A with corresponding patterns of known drugs for which the target proteins to which they bind are known. Based on the pattern matches, the researchers were able to identify eight human receptors and enzymes to which the bacterial molecule could bind. These receptors and enzymes are involved, among other things, in inflammation and pain processes and in the immune system.

Laboratory experiments confirmed that marinopyrrole A did generate measurable interactions with most of the predicted proteins. "Our AI method is able to identify protein targets of natural products with often more than 50 percent reliability, which simplifies the search for new pharmaceutically active agents," Schneider says.

Creating a low-cost alternative

But the work of Schneider's research group is not over. If discoveries about marinopyrrole A's target proteins are to lead to a useful treatment in the future, a molecule that is easy to manufacture is needed. After all, marinopyrrole A - like many other natural substances - has a relatively complicated structure, making laboratory synthesis time-consuming and expensive.

To search for a simpler chemical compound with the same effect, the ETH researchers used another algorithm they designed themselves. This AI program was tasked with being a "virtual chemist" and finding molecules with similar chemical functionality to the natural model despite a different structure. According to the algorithm's constraints, the molecules also had to be able to be made in a maximum of three synthesis steps, to ensure easy and inexpensive production.

New chemical structures with the same effect

To define the synthetic pathway, the software had access to a catalog of more than 200 starting materials, 25,000 purchasable chemical building blocks and 58 established reaction schemes. After each reaction step, the program selected the variants that most closely matched marinopyrrole A in terms of functionality as the starting material for the next step.

In total, the algorithm found 802 suitable molecules based on 334 different scaffolds. The researchers synthesized the top four in the lab and found that they actually behaved very similarly to the natural model. They had a comparable effect on seven of the eight target proteins identified by the algorithm.

Next, the researchers studied the most promising molecule in detail. X-ray structure analyses showed that the compound generated by the counter binds to the active center of a target protein in the same way as known inhibitors of that enzyme. Despite its different structure, the AI-discovered molecule therefore works by the same mechanism.

Effects on pharmaceutical research

"Our work proves that AI algorithms can be employed in a targeted way to design active ingredients with the same effects as natural substances, but with simpler structures," explains Schneider, who adds, "This not only makes it possible to manufacture new drugs, but also places us on the cusp of a potentially fundamental change in medico-chemical research." In other words, the ETH research group's methods make it possible to find drugs that do the same things as existing drugs but are based on different structures. This could make it easier in the future to design new, unpatented molecular structures. The extent to which AI could be used to systematically circumvent patent protection and whether molecules designed by "creative" AI could be patented is currently a hotly debated issue. In any case, the pharmaceutical industry will have to adapt its approach to research to a new regulation.