KAIST Unveils Korean AlphaFold-Style Bio-AI Model K-Fold

Summary

The analysis speed is up to 25 times faster Strategic technology to escape under Big Tech Bio-artificial intelligence (AI), such as Google DeepMind's "Alpha Fold 3," which won the 2024 Nobel Prize in Chemistry, has also emerged in Korea. The speed of predicting protein structure and designing new drug candidates is the fastest in the world. KAIST, the organizer of the 'AI-specific foundation model project' promoted by the government, announced on the 28th that it has unveiled the next-generation bio-AI model 'K-Fold'. The results came less than a year after the start of the project to develop Sovereign AI specialized in the field of scientific research. K-Fold is an AI system that predicts the binding of proteins and drugs needed to develop new drugs. In order to develop a new drug, it is necessary to identify the structure of the causative protein and create a new drug structure that can bind to the protein. The difficulty of grasping the protein structure was the biggest bottleneck in the development of new drugs. AlphaFold won the Nobel Prize in Chemistry earlier because AI solved these bottlenecks. AlphaFold 3, developed by Google DeepMind, accurately predicted the three-dimensional structure of the protein within hours with only the protein amino acid sequence. This is the result of changing the paradigm of new drug development and scientific research. K-Fold is the Korean version of Alpha Fold. BioAI, a strategic asset, is needed for new drug development competitiveness, and bio-specific Sovereign AI developed with its own technology has been secured. Using this, Korea can significantly shorten the time for new drug development and bio-research with its own technology. According to KAIST, the K-Fold has similar performance to the world's best Alpha Fold 3, but is 25 times faster. In our internal analysis, the molecular structure prediction accuracy of K-Fold was close to AlphaFold3 and showed higher performance in some evaluation items. When predicting the next-generation drug platform, PROTAC, K-Fold's prediction was about 33% more accurate than AlphaFold 3. In particular, the prediction success rate was significantly higher than that of Alpha Fold 3 in the work of analyzing how the structure changes when proteins and drugs combine. In the benchmark evaluation, the prediction of changes in activity of the drug was 11% more accurate, and the predictio…