KAIST unveils 'K-Fold,' a sovereign biotech AI model it says is 25 times faster than Google's

A homegrown sovereign biotech AI has emerged to challenge Google DeepMind's AlphaFold3.

KAIST announced Friday that Team KAIST — the institution's research consortium leading the Ministry of Science and ICT's AI-specialized foundation model development project — has developed K-Fold, a next-generation biotech AI model.

The team is led by Professor Kim Woo-youn of the Department of Chemistry, with AI model development handled by professors Hwang Sung-ju and Ahn Sung-su of the Kim Jaechul Graduate School of AI, and protein data construction and validation carried out by professors Oh Byung-ha, Kim Ho-min and Lee Gyu-ri of the Department of Biological Sciences.

At the core of K-Fold is AI-powered prediction of how proteins and drug compounds bind — a critical step in drug development. The model calculates not only the three-dimensional structure of a protein but also where and in what configuration a drug candidate will attach to it, enabling rapid screening of promising new drug candidates.

K-Fold goes beyond predicting the structure of a single protein. It can predict the complex structures formed when multiple biomolecules bind together, including protein-protein and protein-drug candidate interactions as well as combinations involving DNA and RNA.

Its performance approaches world-class levels. In a project milestone evaluation in March, K-Fold's accuracy in predicting molecular complex structures was assessed as comparable to Google DeepMind's AlphaFold3. In an internal performance evaluation the research team conducted in August, K-Fold exceeded existing global models on several benchmarks.

The model showed particularly strong predictive performance in G protein-coupled receptors (GPCRs) and kinases — major drug targets for cancer and other diseases — as well as in targeted protein degradation (TPD), an approach that directly eliminates disease-causing proteins.

K-Fold also delivers a major leap in prediction speed. Conventional protein structure AI models require a complex preprocessing step that involves searching and comparing large volumes of sequence data from similar proteins before any structural calculation can begin. K-Fold eliminates this step by applying a new approach that does not rely on that process, boosting structural prediction speed by up to 25 times compared with existing models.

The research team has also deployed K-Fold as an AI service for use in real research settings. It has been integrated into HyperLab, the multi-agent platform of HITS, a faculty startup at KAIST. Researchers can conduct drug discovery work through a conversational web interface without needing to build dedicated high-performance computing infrastructure or handle complex AI software.

Biotech AI has emerged as a key technology for cutting the enormous time and cost of drug development, drawing fierce competition for dominance among big tech companies and research institutions in the United States, the United Kingdom and China. K-Fold lays the groundwork for sovereign biotech AI by securing core capabilities through domestic technology rather than relying solely on foreign models.

KAIST President Bae Choong-sik said, "National competitiveness in the AI era depends on sovereign AI — the ability to develop and deploy core technologies on our own terms. K-Fold is significant because it combines homegrown AI technology with biotech to challenge world-class performance and connects that capability to a service that can be used in real drug discovery research."

Professor Kim Woo-youn said, "K-Fold was developed not to follow existing models but to apply a new AI architecture that pushes past the limitations of conventional approaches. We aim to grow it into a scientific AI platform that makes world-class biotech AI accessible to any researcher."

Team KAIST plans to release K-Fold free of charge. HyperLab will offer a beta service to domestic and international researchers before rolling out commercial services in stages by the end of this year.