A new generative AI can design molecules from scratch to precisely match corresponding proteins. Image Source: ETH Zurich
Tech Daily News (Reporter: Zhang Mengran) Chemists at ETH Zurich in Switzerland have developed a new artificial intelligence (AI) algorithm that can rapidly and easily design active drug components based on the three-dimensional surface of proteins. This breakthrough, published in the latest issue of "Nature Communications," could potentially revolutionize the way drugs are developed.
The foundation of this method lies in the elucidation of protein three-dimensional structures by chemists over the past decades, as well as the use of computers to search for suitable potential drug molecules. For any known three-dimensional shape of a protein, the algorithm, based on the "lock and key" principle, generates blueprints of potential drug molecules that enhance or inhibit protein activity, which chemists can then synthesize and test in the laboratory.
The new algorithm requires no human intervention; the generative AI can design drug molecules from scratch that match the protein structure.
Furthermore, the algorithm only suggests molecules that interact with specific proteins at desired locations, while hardly interacting with any other proteins. This means that when designing drug molecules, their side effects can be minimized. To create this algorithm, researchers trained the AI model using information on hundreds of thousands of known interactions between chemical molecules and corresponding three-dimensional protein structures.
The research team collaborated with Roche Pharmaceuticals to test the new algorithm and demonstrate the capabilities of AI. They searched for molecules that interact with peroxisome proliferator-activated receptors (PPAR) proteins, which regulate sugar and fatty acid metabolism in the body. Several diabetes drugs currently in use increase PPAR activity, leading cells to absorb more sugar from the blood, thereby lowering blood sugar levels.
Without the need for a lengthy discovery process, AI can immediately design new molecules that also increase PPAR activity, similar to currently available drugs. After producing these molecules in the laboratory, Roche Pharmaceuticals tested them. The results showed that the new molecules were stable and non-toxic from the outset.
