Gwan-Su Yi

Timeline

1. 1984–1993

   Zoology and a KAIST doctorate

   He studies zoology at Seoul National University from 1984 to 1988, then biological engineering at KAIST, where he completes an M.S. in 1990 and a Ph.D. in structural biology in 1993. His doctoral thesis uses NMR to study peptide hormones and the signal sequence of an Escherichia coli ribose-binding protein.

2. 1993–2002

   Fellowships and a proteomics company

   From 1993 to 2001, he is a research fellow at the Protein Engineering Group of the Korea Research Institute of Bioscience, the Magnetic Resonance Group of the Korea Basic Science Institute, the University of North Carolina's Department of Biochemistry and Biophysics, and the University of Toronto's Department of Medical Biophysics. He then works as a senior scientist at Integrative Proteomics (Affinium Pharmaceuticals) in Canada from 2001 to 2002.

3. 2002–2015

   A professorship at KAIST

   In 2002, he joins the Computer Science and Engineering Group at Information and Communications University in Korea as an assistant professor, and he becomes an associate professor there in 2006. He moves to the Department of Bio and Brain Engineering at KAIST as an associate professor in 2009 and becomes a professor in 2015.

4. 2015

   Doctor Kitchen for diabetic diets

   In 2015, Mirae Holdings, a technology holding company created by four science and technology universities including KAIST, creates Doctor Kitchen with technology he developed. The company supplies precooked food and a personalized diet plan that help diabetic patients regulate their diet.

5. 2017

   Drug repositioning by metabolite-likeness

   In 2017, he and Yoon Hyeok Lee publish a BMC Bioinformatics paper on repositioning drugs as enzyme modulators. The work builds a drug-metabolite similarity matrix of 1,861 FDA-approved drugs and 1,110 human intermediary metabolites, and the method outperforms other drug-target prediction methods on known antimetabolite drugs.

6. 2024

   The BASE affinity datasets

   In 2024, he coauthors BASE, the Binding Affinity Similarity Explorer, a web service that provides bias-reduced datasets for deep-learning drug-target affinity prediction. An analysis of eight binding affinity databases with multilayer perceptron models shows that compound features alone predict binding affinity accurately, and BASE addresses this bias by reducing protein similarity between training and test sets.

7. 2026

   GPCRact predicts receptor activation

   In January 2026, he and Hyojin Son publish GPCRact in Briefings in Bioinformatics, a graph neural network that predicts ligand-induced activity of G-protein-coupled receptors by modeling allosteric communication. KAIST announces the model in March 2026 as a way to predict whether a drug candidate activates a protein, a step beyond binding that Google DeepMind's AlphaFold3 does not predict.

8. As of October 2026

   Professor and AI College adjunct

   He is a professor in the Department of Bio and Brain Engineering at KAIST and an adjunct professor at the KAIST AI College. His research area at the AI College is AI-based drug discovery, drawing on scientific machine learning, generative AI and inverse design, and foundation models.