"Design a New Drug for Me": Korea's First Sovereign Bio AI, K-Fold, Debuts

[Financial News] "Design a drug candidate that binds well to this protein."
K-Fold, a bio AI model that designs promising drug candidates for proteins when a researcher makes a request to artificial intelligence (AI), has been developed with Korea's own technology.
The Korea Advanced Institute of Science and Technology (KAIST) said on the 28th that it formed Team KAIST as the lead institution for the Ministry of Science and ICT's "AI-specialized foundation model project" and unveiled its next-generation bio AI model, K-Fold.
Team KAIST is led by Professor Kim Woo-yeon of the Department of Chemistry. Researchers from his group and from the groups of Professors Hwang Seong-ju and Ahn Seong-su at the Kim Jaechul Graduate School of AI at KAIST (KAIST AI) are developing the AI model, while Professors Oh Byung-ha, Kim Ho-min, and Lee Gyu-ri of the Department of Biological Sciences are responsible for building and validating protein data. HITS, a faculty startup at KAIST, has turned K-Fold into a service linked with HyperLab, a web-based AI research platform, so that researchers can use it in practice. The Korea Pharmaceutical and Bio-Pharma Manufacturers Association and the Korea Biotechnology Industry Organization (KBIO) will handle the spread of K-Fold's achievements and promote its use in industry.
K-Fold's biggest feature is that it uses AI to predict the "binding" between proteins and drugs needed for new drug development. It not only predicts the three-dimensional structure of proteins, but also calculates in advance where and how a drug candidate will bind to a protein, helping researchers quickly identify promising candidates. It can also predict the structures formed when various biomolecules, including protein-protein, protein-drug candidate, and DNA-RNA interactions, come together.
Its performance is also world-class. In a March interim evaluation, K-Fold's accuracy in predicting molecular complex structures was assessed as approaching that of Google DeepMind's AlphaFold 3. In a performance test conducted by the research team in August, it also outperformed existing global models in some evaluation categories.
In particular, it showed strong performance in predicting how drugs bind to and act on target proteins in areas such as G protein-coupled receptors (GPCRs) and protein kinases, which are major drug targets for diseases including cancer, as well as targeted protein degradation (TPD), a new drug technology that directly removes disease-causing proteins.
It also improved structural prediction speed by up to 25 times compared with existing models. That means more drug candidates can be examined in the same amount of time.
The research team did not stop at developing K-Fold. It also turned the model into an AI research service that researchers can use through web-based conversation. By integrating K-Fold into HyperLab, HITS's multi-agent platform, the team made it possible to use the system without building a separate high-performance computer or handling complex AI programs. Instead of moving between multiple programs to calculate structures and analyze results, researchers can simply tell the AI their research goals, and it carries out the necessary analysis and design step by step.
Kim Woo-yeon, a professor in KAIST's Department of Chemistry, said, "K-Fold was developed not to follow existing models, but to overcome the limitations of conventional methods by applying a new AI architecture." He added, "We will develop world-class bio AI into a science AI platform that any researcher can use."
The Team KAIST consortium plans to distribute K-Fold free of charge, while HyperLab will offer a beta service to researchers in Korea and abroad before gradually expanding commercial services within this year.

[email protected] Yeon Jian Reporter