Prospective discovery of LasB inhibitors by integrating generative design, free energy ranking, and experimental validation
This study presents a fully open-source computational-experimental workflow that integrates generative molecular design with rigorous free energy calculations to successfully identify three potent, stable nanomolar inhibitors of the *Pseudomonas aeruginosa* virulence factor LasB, demonstrating a scalable framework for accelerating early-stage drug discovery.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer
Finding a new medicine is like searching for a single, specific key in a library containing billions of locks. The process is slow, expensive, and often frustrating, because chemists must build and test millions of potential candidates by hand to see which ones actually fit. For decades, scientists have tried to speed this up with computers, using them to predict which molecules might work before a single atom is physically assembled. However, these computer predictions have often been too rough to trust on their own, leading researchers to waste time synthesizing compounds that turn out to be useless. The challenge has been to create a system that is both fast enough to explore vast new chemical territories and precise enough to distinguish a true winner from a near-miss.
In a recent study, researchers have successfully combined two powerful computer techniques to solve this problem, creating a workflow that moves from digital imagination to physical reality with remarkable efficiency. They focused their efforts on a dangerous bacterium called Pseudomonas aeruginosa, a common cause of severe hospital infections that is becoming increasingly difficult to treat with existing antibiotics. This bacterium relies on a specific protein, a molecular machine known as LasB, to break down human tissues and spread infection. Because this protein is secreted outside the bacterial cell, it is easier to reach with a drug than targets hidden inside the bacteria's tough outer shell. The goal was to design a new molecule that could lock onto this protein and stop it from working, effectively disarming the bacteria without killing it, a strategy that might reduce the chance of the bacteria developing resistance.
The team began with three known molecules that already showed some ability to bind to the LasB protein, but they were not perfect. Using a computer program designed to generate new molecular structures, they started with these three seeds and asked the software to imagine thousands of variations. The program did not just shuffle atoms randomly; it was guided by rules that favored molecules likely to be effective drugs and easy to make in a lab. Over several rounds of digital evolution, the software generated nearly 40,000 unique candidate molecules. From this massive digital library, the researchers selected a small, diverse group of 50 compounds that looked the most promising based on their shape and chemical properties.
To ensure these 50 candidates were truly worth the cost of building them, the team applied a second, more rigorous computer test. This step involved simulating the physical forces between the drug candidate and the protein with extreme precision, calculating exactly how tightly they would hold onto each other. This method, known as free energy calculation, is computationally expensive but far more accurate than standard screening methods. The results of this simulation allowed the team to rank the 50 candidates and pick the very best ones for physical synthesis. They built six of these top-ranked molecules in the laboratory and tested them against the bacterial protein.
The outcome was a clear success for this integrated approach. Three of the newly created molecules proved to be potent inhibitors, stopping the bacterial protein with high efficiency at concentrations as low as 22 to 53 nanomolar. These compounds were not only powerful but also showed excellent stability in blood and metabolic tests, meaning they would likely survive long enough in the body to do their job. While the new drugs showed some toxicity to human cells in a dish, their strong ability to target the bacterial protein suggests a viable path forward for further development. The study demonstrates that by linking generative design, which creates new ideas, with rigorous physics-based ranking, which tests those ideas with high precision, scientists can navigate the vast ocean of chemical possibilities and find valuable treasures without getting lost.
The researchers also learned important lessons about the limits of their method. They found that while the computer predictions were excellent at ranking which molecules were better than others, they sometimes overestimated exactly how strong the bond would be. This discrepancy likely occurred because the computer had to guess the initial position of the new molecules before running the detailed simulation, and a small error in that guess could throw off the final calculation. Despite this, the system worked well enough to identify three completely new, highly effective inhibitors from a pool of thousands, proving that this digital-first strategy can guide experimental work more effectively than traditional trial-and-error. The study concludes that this approach offers a scalable, open-source framework that other scientists can use to tackle similar challenges, potentially accelerating the discovery of medicines for other difficult diseases.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.