Contagious Science: Artificial Intelligence Has Developed an Extremely Strong Medical Glue That Also Works in an Aquatic Environment

New research shows how a combination of natural protein pools and machine learning led to the development of a hydrogel with extraordinary adhesive properties, which may help in medicine and advanced materials in the future.

One of the biggest challenges in materials science today is creating adhesives that work in an aqueous environment. Sounds boring? Tell that to a woman whose amniotic sac has ruptured, and the fluid that keeps her fetus alive is leaking out. Or to a person who's been in a car accident and is suffering from internal bleeding that doctors just can't stop. Or to a child whose tendon has ruptured.

All of these cases – and many others in medicine – require the use of an adhesive that can work inside the body. But such adhesives are rare, few in number, and usually very weak.

So what is needed? I've already said: invent better adhesives that work in an aquatic environment. But between what is needed and what is possible there is a very long road of trial and error and Sisyphean labor in laboratories. All this work is even more frustrating because we know that nature has already solved the problem. Clams and snails, for example, cling tightly to wet rocks and manage to stay firmly attached to them despite the waves that constantly beat against them. But human attempts to mimic their chemical composition in the laboratory have lasted decades and have mostly yielded disappointments.

Then, last year, a study was published in the journal Nature Which showed how the rules of the game in laboratories are changing. 

The researchers used artificial intelligence to sift through nature’s “patent library.” They fed a computer a vast database of nearly 25,000 different adhesive proteins from across the animal kingdom. Smart algorithms combed through this mountain of biological information, extracting insights into the recipes that make certain proteins more or less sticky. 

Now that they had a better understanding of the subject, the researchers created 180 “recipes” for a gel-like material (hydrogel). They tested them in the lab, and fed the results into a machine learning model. At this point, they let the artificial intelligence do what the human brain has difficulty with: spot recurring patterns in a wealth of data. The system analyzed the information and came up with a winning recipe for an entirely new material, designed entirely from scratch – out of thin air. But does this material really work? To test it, the researchers implemented the recipe in the lab, obtained a new adhesive, and conducted a series of experiments on it.

And when they finished the experiments, they probably invited all the lab members for a beer at the bar, because they had a winning glue in hand.

The new glue they produced with the help of artificial intelligence broke natural world records. It demonstrated an adhesive force ten times stronger than any other hydrogel glue that works in an aquatic environment that we have known so far. And it is even reusable and works in saltwater. We know this because the researchers demonstrated its capabilities to glue… a duck to rocks in the sea. Don’t worry, it’s a rubber duck. And yet, it stuck well to the rocks thanks to the new glue, and bravely dealt with all the waves that the sea made sure to throw in its direction.

The implications of this breakthrough are doubly exciting. In the short term, we have (maybe) a life-saving “superglue.” Researchers haven’t tested this glue inside the body, but if it gets the necessary clinical approvals, it could be used in operating rooms to seal bleeding organs in seconds without the need for stitches. Or we could attach tiny medical sensors to internal tissues. A new world of intracorporeal therapies would open up for us.

Is that what will happen? Maybe. I'm always wary of betting on just one substance. Biology is a complex science, and it's impossible to know which substance will prove itself in clinical trials, and which will fail. Still, there is potential for success, and that's encouraging.

The real, and more optimistic, news concernshow We are inventing. We are witnessing a turning point in the history of science. By combining the enormous biological data with the computational power of artificial intelligence, cycles of innovation, development, and invention that once required decades of frustrated student lab work are now being reduced to a matter of months.

The use of AI in the laboratory will continue to expand, both because it is delivering clear successes and because it is increasingly easy to use today. In fact, the most advanced models are already able to tell researchers how to use them for these purposes. And while one must be careful not to blindly rely on AI recommendations, this means that the level of expertise needed to benefit from them is rapidly decreasing.

And so, we are entering a period where the pace of discovery, development, and invention will improve dramatically. In all areas. Today it's strong underwater glue. Tomorrow it could be self-degrading plastics, self-healing artificial skin, or robotic muscles. 

Nature has never been more open to analysis and experimentation. Artificial intelligence helps us take hundreds of millions of years of evolution and produce new insights and drugs at breakneck speed, leading to groundbreaking medical treatments, advanced energy technologies, and much more.

Every field – medicine, agriculture, engineering – is set to improve significantly in the coming years thanks to the ability to develop and invent inventions faster and better.

Now we just have to hope that we know how to use all these inventions for the better.

It's time to remind you that on May 18th I will speak At the Science Abroad conference in New York About the future of science and research. You are welcome to come and meet me there!