The woodblock print 'Under the Wave off Kanagawa' created by Katsushika Hokusai in 1831 is a painting that symbolizes Japan. One of the Thirty-Six Views of Mount Fuji, it depicts the distant Mount Fuji, capped with white snow, seen between giant waves. This painting also influenced contemporary Impressionist painters such as Manet, Monet, and Van Gogh.
Hokusai's wave painting was featured in the international journal 'Nature.' It is an illustration of a rubber duck strongly adhered to a rock on a beach with towering waves. Jian Ping Gong, a professor at Hokkaido University in Japan, said on the 7th, "An underwater adhesive created by AI fixed the rubber duck to the rock amid fierce waves."
◇AI learns 25,000 types of natural adhesives
Professor Gong's research team developed a medical hydrogel adhesive. Hydrogel is a soft polymer substance that contains a lot of water, like jelly or pudding. The human body is composed of 70% water. Medical adhesives must bond firmly even in water. Scientists sought answers from nature.
Mussels firmly attach themselves to rocks even in crashing waves. Sandworms collect surrounding sand or shell fragments in seawater to build solid cylindrical homes. Scientists developed medical adhesives that mimic strong natural adhesives found in water to suture surgical sites and bond broken bones.
The problem is that learning from nature takes a lot of time and expense. In Korea, a surgical adhesive mimicking mussel protein was commercialized after over 20 years of research. AI doesn't have this issue; it can synthesize natural adhesives quickly as long as data is provided.
Researchers at Hokkaido University gathered information on 24,707 natural adhesive substances from the National Center for Biotechnology Information (NCBI). This included information on 3,822 organisms that produce adhesive substances, ranging from bacteria and viruses to mussels and sandworms. Proteins are made up of linked amino acids. Natural adhesive proteins consist of 300 to 500 amino acids.
The research team designed new adhesive proteins using AI's machine learning and deep learning. Machine learning is an AI technology that learns large amounts of data and finds methods independently without prior programming. Deep learning is a type of deep machine learning that mimics human neural structures.
AI identified motifs, specific amino acid sequences that are the origins of adhesive power in natural substances. Through this, it created 180 new underwater adhesives. The research team synthesized the adhesives designed by AI and tested their adhesive strength, then used this information to retrain AI to create the optimal underwater adhesive. One of these, R1-max, a hydrogel adhesive, firmly fixed the rubber duck to the rock amid crashing waves. R2-max sealed a 20 mm diameter hole that formed at the bottom of a 3 m long pipe filled with water, with no leaks for five months.
◇A stamp size can support the weight of an adult
Professor Gong's research team stated that the hydrogel adhesive created by AI has an adhesive strength exceeding 1 MPa (megapascal), making it the best underwater adhesive currently available. 1 MPa can withstand a force of 10 kg per square centimeter. The research team noted, "If the hydrogel adhesive is made to the size of a stamp, measuring 2.5 cm on each side, it could theoretically support the weight of an adult weighing 63 kg."
AI has previously been utilized in materials design, but most of the time, it was used to create solid inorganic materials with uniform structure and properties. Natural underwater adhesives are complex organic polymer substances influenced by various factors, making them difficult to design with AI.
Furthermore, there was a lack of data for AI to learn from. The research team optimized AI design this time by training it with large-scale information and then synthizing the adhesives as AI designed them, feeding the experimental results back to AI for further learning.
Laura Russo, a professor at the University of Milan-Bicocca in Italy, remarked in a commentary paper published in Nature that "this research shows that AI is actively transforming the way scientists conduct their research, rather than merely remaining in the exploratory phase in materials science." She added, "The hydrogel adhesive, which strongly adheres to irregular and wet surfaces, can be used in various medical devices, such as medical prosthetics and wearable biosensors."
References
Nature (2025), DOI: https://doi.org/10.1038/s41586-025-09269-4