%44%72%75%67%20%64%65%76%65%6c%6f%70%6d%65%6e%74
-
KAIST Develops AI That Automatically Designs Optimal Drug Candidates for Cancer-Targeting Mutations
< (From left) Ph.D candidate Wonho Zhung, Ph.D cadidate Joongwon Lee , Prof. Woo Young Kim , Ph.D candidate Jisu Seo >
Traditional drug development methods involve identifying a target protin (e.g., a cancer cell receptor) that causes disease, and then searching through countless molecular candidates (potential drugs) that could bind to that protein and block its function. This process is costly, time-consuming, and has a low success rate. KAIST researchers have developed an AI model that, using only information about the target protein, can design optimal drug candidates without any prior molecular data—opening up new possibilities for drug discovery.
KAIST (President Kwang Hyung Lee) announced on the 10th that a research team led by Professor Woo Youn Kim in the Department of Chemistry has developed an AI model named BInD (Bond and Interaction-generating Diffusion model), which can design and optimize drug candidate molecules tailored to a protein’s structure alone—without needing prior information about binding molecules. The model also predicts the binding mechanism (non-covalent interactions) between the drug and the target protein.
The core innovation of this technology lies in its “simultaneous design” approach. Previous AI models either focused on generating molecules or separately evaluating whether the generated molecule could bind to the target protein. In contrast, this new model considers the binding mechanism between the molecule and the protein during the generation process, enabling comprehensive design in one step. Since it pre-accounts for critical factors in protein-ligand binding, it has a much higher likelihood of generating effective and stable molecules. The generation process visually demonstrates how types and positions of atoms, covalent bonds, and interactions are created simultaneously to fit the protein’s binding site.
<Figure 1. Schematic of the diffusion model developed by the research team, which generates molecular structures and non-covalent interactions based on protein structures. Starting from a noise distribution, the model gradually removes noise (via reverse diffusion) to restore the atom positions, types, covalent bond types, and interaction types, thereby generating molecules. Interacting patterns are extracted from prior knowledge of known binding molecules or proteins, and through an inpainting technique, these patterns are kept fixed during the reverse diffusion process to guide the molecular generation.>
Moreover, this model is designed to meet multiple essential drug design criteria simultaneously—such as target binding affinity, drug-like properties, and structural stability. Traditional models often optimized for only one or two goals at the expense of others, but this new model balances various objectives, significantly enhancing its practical applicability.
The research team explained that the AI operates based on a “diffusion model”—a generative approach where a structure becomes increasingly refined from a random state. This is the same type of model used in AlphaFold 3, the 2024 Nobel Chemistry Prize-winning tool for protein-ligand structure generation, which has already demonstrated high efficiency.
Unlike AlphaFold 3, which provides spatial coordinates for atom positions, this study introduced a knowledge-based guide grounded in actual chemical laws—such as bond lengths and protein-ligand distances—enabling more chemically realistic structure generation.
<Figure 2. (Left) Target protein and the original bound molecule; (Right) Examples of molecules designed using the model developed in this study. The values for protein binding affinity (Vina), drug-likeness (QED), and synthetic accessibility (SA) are shown at the bottom.>
Additionally, the team applied an optimization strategy where outstanding binding patterns from prior results are reused. This allowed the model to generate even better drug candidates without additional training. Notably, the AI successfully produced molecules that selectively bind to the mutated residues of EGFR, a cancer-related target protein.
This study is also meaningful because it advances beyond the team’s previous research, which required prior input about the molecular conditions for the interaction pattern of protein binding.
Professor Woo Youn Kim commented that “the newly developed AI can learn and understand the key features required for strong binding to a target protein, and design optimal drug candidate molecules—even without any prior input. This could significantly shift the paradigm of drug development.” He added, “Since this technology generates molecular structures based on principles of chemical interactions, it is expected to enable faster and more reliable drug development.”
Joongwon Lee and Wonho Zhung, PhD students in the Department of Chemistry, participated as co-first authors of this study. The research results were published in the international journal Advanced Science (IF = 14.1) on July 11.
● Paper Title: BInD: Bond and Interaction-Generating Diffusion Model for Multi-Objective Structure-Based Drug Design
● DOI: 10.1002/advs.202502702
This research was supported by the National Research Foundation of Korea and the Ministry of Health and Welfare.
2025.08.12 View 3720 -
“One Experiment Is All It Takes”: KAIST Team Revolutionizes Drug Interaction Testing, Replacing 60,000 Studies
A groundbreaking new method developed by researchers at KAIST and Chungnam National University could drastically streamline drug interaction testing — replacing dozens of traditional experiments with just one.
The research, led by Professor Jae Kyoung Kim of KAIST Department of Mathematical Sciences & IBS Biomedical Mathematics Group and Professor Sang Kyum Kim of Chungnam National University's College of Pharmacy, introduces a novel analysis technique called 50-BOA, published in Nature Communications on June 5, 2025.
< Photo 1. (From left) Professor Sang Kyum Kim (Chungnam National University College of Pharmacy, co-corresponding author), Dr. Yun Min Song (IBS Biomedical Mathematics Group, formerly KAIST Department of Mathematical Sciences, co-first author), undergraduate student Hyeong Jun Jang (KAIST, co-first author), Professor Jae Kyoung Kim (KAIST and IBS Biomedical Mathematics Group, co-corresponding author) (Top left in the bubble) Professor Hwi-yeol Yun (Chungnam National University College of Pharmacy, co-author) >
For decades, scientists have had to repeat drug inhibition experiments across a wide range of concentrations to estimate inhibition constants — a process seen in over 60,000 scientific publications. But the KAIST-led team discovered that a single, well-chosen inhibitor concentration can yield even more accurate results.
< Figure 1. Graphical summary of 50-BOA. 50-BOA improves the accuracy and efficiency of inhibition constant estimation by using only a single inhibitor concentration instead of the traditionally used method of employing multiple inhibitor concentrations. >
“This approach challenges long-standing assumptions in experimental pharmacology,” says Prof. Kim. “It shows how mathematics can fundamentally redesign life science experiments.”
By mathematically analyzing the sources of error in conventional methods, the team found that over half the data typically collected adds no value or even skews results. Their new method not only cuts experimental effort by over 75%, but also enhances reproducibility and accuracy.
To help researchers adopt the method quickly, the team developed a user-friendly tool that takes simple Excel files as input, now freely available on GitHub:
☞ https://github.com/Mathbiomed/50-BOA
< Figure 2. The MATLAB and R package of 50-BOA at GitHub >
The work holds promise for faster and more reliable drug development, especially in assessing potential interactions in combination therapies. The U.S. FDA already emphasizes the importance of accurate enzyme inhibition assessment during early-stage drug evaluation — and this method could soon become a new gold standard.
2025.06.16 View 6829 -
A Mathematical Model Reveals Long-Distance Cell Communication Mechanism
How can tens of thousands of people in a large football stadium all clap together with the same beat even though they can only hear the people near them clapping?
A combination of a partial differential equation and a synthetic circuit in microbes answers this question. An interdisciplinary collaborative team of Professor Jae Kyoung Kim at KAIST, Professor Krešimir Josić at the University of Houston, and Professor Matt Bennett at Rice University has identified how a large community can communicate with each other almost simultaneously even with very short distance signaling. The research was reported at Nature Chemical Biology.
Cells often communicate using signaling molecules, which can travel only a short distance. Nevertheless, the cells can also communicate over large distances to spur collective action. The team revealed a cell communication mechanism that quickly forms a network of local interactions to spur collective action, even in large communities.
The research team used an engineered transcriptional circuit of combined positive and negative feedback loops in E. coli, which can periodically release two types of signaling molecules: activator and repressor. As the signaling molecules travel over a short distance, cells can only talk to their nearest neighbors. However, cell communities synchronize oscillatory gene expression in spatially extended systems as long as the transcriptional circuit contains a positive feedback loop for the activator.
Professor Kim said that analyzing and understanding such high-dimensional dynamics was extremely difficult. He explained, “That’s why we used high-dimensional partial differential equation to describe the system based on the interactions among various types of molecules.” Surprisingly, the mathematical model accurately simulates the synthesis of the signaling molecules in the cell and their spatial diffusion throughout the chamber and their effect on neighboring cells.
The team simplified the high-dimensional system into a one-dimensional orbit, noting that the system repeats periodically. This allowed them to discover that cells can make one voice when they lowered their own voice and listened to the others. “It turns out the positive feedback loop reduces the distance between moving points and finally makes them move all together. That’s why you clap louder when you hear applause from nearby neighbors and everyone eventually claps together at almost the same time,” said Professor Kim.
Professor Kim added, “Math is a powerful as it simplifies complex thing so that we can find an essential underlying property. This finding would not have been possible without the simplification of complex systems using mathematics."
The National Institutes of Health, the National Science Foundation, the Robert A. Welch Foundation, the Hamill Foundation, the National Research Foundation of Korea, and the T.J. Park Science Fellowship of POSCO supported the research.
(Figure: Complex molecular interactions among microbial consortia is simplified as interactions among points on a limit cycle (right).)
2019.10.15 View 35742 -
Mathematical Modeling Makes a Breakthrough for a New CRSD Medication
PhD Candidate Dae Wook Kim (Left) and Professor Jae Kyoung Kim (Right)
- Systems approach reveals photosensitivity and PER2 level as determinants of clock-modulator efficacy -
Mathematicians’ new modeling has identified major sources of interspecies and inter-individual variations in the clinical efficacy of a clock-modulating drug: photosensitivity and PER2 level. This enabled precision medicine for circadian disruption.
A KAIST mathematics research team led by Professor Jae Kyoung Kim, in collaboration with Pfizer, applied a combination of mathematical modeling and simulation tools for circadian rhythms sleep disorders (CRSDs) to analyze the animal data generated by Pfizer. This study was reported in Molecular Systems Biology as the cover article on July 8.
Pharmaceutical companies have conducted extensive studies on animals to determine the candidacy of this new medication. However, the results of animal testing do not always translate to the same effects in human trials. Furthermore, even between humans, efficacy differs across individuals depending on an individual’s genetic and environmental factors, which require different treatment strategies.
To overcome these obstacles, KAIST mathematicians and their collaborators developed adaptive chronotherapeutics to identify precise dosing regimens that could restore normal circadian phase under different conditions.
A circadian rhythm is a 24-hour cycle in the physiological processes of living creatures, including humans. A biological clock in the hypothalamic suprachiasmatic nucleus in the human brain sets the time for various human behaviors such as sleep.
A disruption of the endogenous timekeeping system caused by changes in one’s life pattern leads to advanced or delayed sleep-wake cycle phase and a desynchronization between sleep-wake rhythms, resulting in CRSDs. To restore the normal timing of sleep, timing of the circadian clock could be adjusted pharmacologically.
Pfizer identified PF-670462, which can adjust the timing of circadian clock by inhibiting the core clock kinase of the circadian clock (CK1d/e). However, the efficacy of PF-670462 significantly differs between nocturnal mice and diurnal monkeys, whose sleeping times are opposite.
The research team discovered the source of such interspecies variations in drug response by performing thousands of virtual experiments using a mathematical model, which describes biochemical interactions among clock molecules and PF-670462. The result suggests that the effect of PF-670462 is reduced by light exposure in diurnal primates more than in nocturnal mice. This indicates that the strong counteracting effect of light must be considered in order to effectively regulate the circadian clock of diurnal humans using PF-670462.
Furthermore, the team also found the source of inter-patients variations in drug efficacy using virtual patients whose circadian clocks were disrupted due to various mutations. The degree of perturbation in the endogenous level of the core clock molecule PER2 affects the efficacy.
This explains why the clinical outcomes of clock-modulating drugs are highly variable and certain subtypes are unresponsive to treatment. Furthermore, this points out the limitations of current treatment strategies tailored to only the patient’s sleep and wake time but not to the molecular cause of sleep disorders.
PhD candidate Dae Wook Kim, who is the first author, said that this motivates the team to develop an adaptive chronotherapy, which identifies a personalized optimal dosing time of day by tracking the sleep-wake up time of patients via a wearable device and allows for a precision medicine approach for CRSDs.
Professor Jae Kyoung Kim said, "As a mathematician, I am excited to help enable the advancement of a new drug candidate, which can improve the lives of so many patients. I hope this result promotes more collaborations in this translational research.”
This research was supported by a Pfizer grant to KAIST (G01160179), the Human Frontiers Science Program Organization (RGY0063/2017), and a National Research Foundation (NRF) of Korea Grant (NRF-2016 RICIB 3008468 and NRF-2017-Fostering Core Leaders of the Future Basic Science Program/ Global Ph.D. Fellowship Program).
Figure 1. Interspecies and Inter-patients Variations in PF-670462 Efficacy
Figure 2. Journal Cover Page
Publication:
Dae Wook Kim, Cheng Chang, Xian Chen, Angela C Doran, Francois Gaudreault, Travis Wager, George J DeMarco, and Jae Kyoung Kim. 2019. Systems approach reveals photosensitivity and PER2 level as determinants of clock-modulator efficacy. Molecular Systems Biology. EMBO Press, Heidelberg, Germany, Vol. 15, Issue No. 7, Article, 16 pages. https://doi.org/10.15252/msb.20198838
Profile: Prof. Jae Kyoung Kim, PhD
jaekkim@kaist.ac.kr
http://mathsci.kaist.ac.kr/~jaekkim
Associate Professor
Department of Mathematical Sciences
Korea Advanced Institute of Science and Technology (KAIST)
http://kaist.ac.kr Daejeon 34141, Korea
Profile: Dae Wook Kim, PhD Candidate
0308kdo@kaist.ac.kr
http://mathsci.kaist.ac.kr/~jaekkim
PhD Candidate
Department of Mathematical Sciences
Korea Advanced Institute of Science and Technology (KAIST)
http://kaist.ac.kr Daejeon 34141, Korea
Profile: Dr. Cheng Chang, PhD
cheng.chang@pfizer.com
Associate Director of Clinical Pharmacology
Clinical Pharmacology, Global Product Development
Pfizer
https://www.pfizer.com/ Groton 06340, USA
(END)
2019.07.09 View 41240 -
Professor Jae Kyoung Kim Receives the 2017 HSFP Award
The Human Frontier Science Program (HSFP), one of the most competitive research grants in life sciences, has funded researchers worldwide across and beyond the field since 1990. Each year, the program selects a handful of recipients who push the envelope of basic research in biology to bring breakthroughs from novel approaches. Among its 7,000 recipients thus far, 26 scientists have received the Nobel Prize. For that reason, HSFP grants are often referred to as “Nobel Prize Grants.”
Professor Jae Kyoung Kim of the Mathematical Sciences Department at KAIST and his international collaborators, Professor Robert Havekes from the University of Groningen, the Netherlands, Professor Sara Aton from the University of Michigan in Ann Arbor, the United States, and Professor Matias Zurbriggen from the University of Düsseldorf, Germany, won the Young Investigator Grants of the 2017 HSFP.
The 30 winning teams of the 2017 competition (in 9 Young Investigator Grants and 21 Program Grants) went through a rigorous year-long review process from a total of 1,073 applications submitted from more than 60 countries around the world. Each winning team will receive financial support averaging 110,000-125,000 USD per year for three years.
Although Professor Kim was trained as a mathematician, he has extended his research focus into biological sciences and attempted to solve some of the most difficult problems in biology by employing mathematical theories and applications including nonlinear dynamics, stochastic process, singular perturbation, and parameter estimation.
The project that won the Young Investigator Grants was a study on how a molecular circadian clock may affect sleep-regulated neurophysiology in mammals. Physiological and metabolic processes such as sleep, blood pressure, and hormone secretion exhibit circadian rhythms in mammals. Professor Kim used mathematical modeling and analysis to explain that the mammalian circadian clock is a hierarchical system, in which the master clock in the superchiasmatic nucleus, a tiny region in the brain that controls circadian rhythms, functions as a pacemaker and synchronizer of peripheral clocks to generate coherent systematic rhythms throughout the body.
Professor Kim said, “The mechanisms of our neuronal and hormonal activities regulating many of our bodily functions over a 24-hour cycle are not yet fully known. We go to sleep every night, but do not really know how it affects our brain functions. I hope my experience in mathematics, along with insights from biologists, can find meaningful answers to some of today’s puzzling problems in biological sciences, for example, revealing the complexities of our brains and showing how they work.”
“In the meantime, I hope collaborations between the fields of mathematics and biology, as yet a rare phenomenon in the Korean scientific community, will become more popular in the near future.”
Professor Kim received his doctoral degree in Applied and Interdisciplinary Mathematics in 2013 from the University of Michigan and joined KAIST in 2015. He has published numerous articles in reputable science journals such as Science, Molecular Cell, Proceedings of the National Academy of Sciences, and Nature Communications.
Both the Program Grants and Young Investigator Grants support international teams with members from at least two countries for innovative and creative research. This year, the Program Grants were awarded to research topics ranging from the evolution of counting and the role of extracellular vesicles in breast cancer bone metastasis to the examination of obesity from a mechanobiological point of view.
The Young Investigator Grants are limited to teams that established their independent research within the last five years and received their doctoral degrees within the last decade. Besides Professor Kim’s study, such topics as the use of infrasound for navigation by seabirds and protein formation in photochemistry and photophysics were awarded in 2017.
Full lists of the 2017 HFSP winners are available at: http://www.hfsp.org/awardees/newly-awarded.
About the Human Frontier Science Program (HFSP):
The HFSP is a research funding program implemented by the International Human Frontier Science Program (HFSPO) based in Strasbourg, France. It promotes intercontinental collaboration and training in cutting-edge, interdisciplinary research specializing in life sciences. Founded in 1989, the HFSPO consists of the European Union and 14 other countries including the G7 nations and South Korea.
2017.03.21 View 15317 -
System Approach Using Metabolite Structural Similarity Toward TOM Suggested
A Korean research team at KAIST suggests that a system approach using metabolite structural similarity helps to elucidate the mechanisms of action of traditional oriental medicine.
Traditional oriental medicine (TOM) has been practiced in Asian countries for centuries, and is gaining increasing popularity around the world. Despite its efficacy in various symptoms, TOM has been practiced without precise knowledge of its mechanisms of action. Use of TOM largely comes from empirical knowledge practiced over a long period of time. The fact that some of the compounds found in TOM have led to successful modern drugs such as artemisinin for malaria and taxol (Paclitaxel) for cancer has spurred modernization of TOM.
A research team led by Sang-Yup Lee at KAIST has focused on structural similarities between compounds in TOM and human metabolites to help explain TOM’s mechanisms of action. This systems approach using structural similarities assumes that compounds which are structurally similar to metabolites could affect relevant metabolic pathways and reactions by biosynthesizing structurally similar metabolites.
Structural similarity analysis has helped to identify mechanisms of action of TOM. This is described in a recent study entitled “A systems approach to traditional oriental medicine,” published online in Nature Biotechnology on March 6, 2015. In this study, the research team conducted structural comparisons of all the structurally known compounds in TOM and human metabolites on a large-scale. As a control, structures of all available approved drugs were also compared against human metabolites. This structural analysis provides two important results. First, the identification of metabolites structurally similar to TOM compounds helped to narrow down the candidate target pathways and reactions for the effects from TOM compounds. Second, it suggested that a greater fraction of all the structurally known TOM compounds appeared to be more similar to human metabolites than the approved drugs. This second finding indicates that TOM has a great potential to interact with diverse metabolic pathways with strong efficacy. This finding, in fact, shows that TOM compounds might be advantageous for the multitargeting required to cure complex diseases. “Once we have narrowed down candidate target pathways and reactions using this structural similarity approach, additional in silico tools will be necessary to characterize the mechanisms of action of many TOM compounds at a molecular level,” said Hyun Uk Kim, a research professor at KAIST.
TOM’s multicomponent, multitarget approach wherein multiple components show synergistic effects to treat symptoms is highly distinctive. The researchers investigated previously observed effects recorded since 2000 of a set of TOM compounds with known mechanisms of action. TOM compounds’ synergistic combinations largely consist of a major compound providing the intended efficacy to the target site and supporting compounds which maximize the efficacy of the major compound. In fact, such combination designs appear to mirror the Kun-Shin-Choa-Sa design principle of TOM.
That principle, Kun-Shin-Choa-Sa (君臣佐使 or Jun-Chen-Zuo-Shi in Chinese) literally means “king-minister-assistant-ambassador.” In ancient East Asian medicine, treating human diseases and taking good care of the human body are analogous to the politics of governing a nation. Just as good governance requires that a king be supported by ministers, assistants and/or ambassadors, treating diseases or good care of the body required the combined use of herbal medicines designed based on the concept of Kun-Shin-Choa-Sa. Here, the Kun (king or the major component) indicates the major medicine (or herb) conveying the major drug efficacy, and is supported by three different types of medicines: the Shin (minister or the complementary component) for enhancing and/or complementing the efficacy of the Kun, Choa (assistant or the neutralizing component) for reducing any side effects caused by the Kun and reducing the minor symptoms accompanying major symptom, and Sa (ambassador or the delivery/retaining component) which facilitated the delivery of the Kun to the target site, and retaining the Kun for prolonged availability in the cells.
The synergistic combinations of TOM compounds reported in the literature showed four different types of synergisms: complementary action (similar to Kun-Shin), neutralizing action (similar to Kun-Choa), facilitating action or pharmacokinetic potentiation (both similar to Kun-Choa or Kun-Sa). Additional structural analyses for these compounds with synergism show that they appeared to affect metabolism of amino acids, co-factors and vitamins as major targets.
Professor Sang Yup Lee remarks, “This study lays a foundation for the integration of traditional oriental medicine with modern drug discovery and development. The systems approach taken in this analysis will be used to elucidate mechanisms of action of unknown TOM compounds which will then be subjected to rigorous validation through clinical and in silico experiments.”
Sources: Kim, H.U. et al. “A systems approach to traditional oriental medicine.” Nature Biotechnology. 33: 264-268 (2015).
This work was supported by the Bio-Synergy Research Project (2012M3A9C4048759) of the Ministry of Science, ICT and Future Planning through the National Research Foundation. This work was also supported by the Novo Nordisk Foundation.
The picture below presents the structural similarity analysis of comparing compounds in traditional oriental medicine and those in all available approved drugs against the structures of human metabolites.
2015.03.09 View 13264 -
Light Driven Drug-Enzyme Reaction Catalytic Platform Developed
Low Cost Dye Used, Hope for Future Development of High Value Medicinal Products to Treat Cardiovascular Disease and Gastric Ulcers
A KAIST research team from the Departments of Materials Science and Engineering and of Chemical and Biomolecular Engineering, led respectively by Professors Chan Beum Park and Ki Jun Jeong, has developed a new reaction platform to induce drug-enzyme reaction using light.
The research results were published in the journal Angewandte Chemie, International Edition, as the back cover on 12 January 2015.
Applications of this technology may enable production of high value products such as medicine for cardiovascular disease and gastric ulcers, for example Omeprazole, using an inexpensive dye.
Cytochrome P450 is an enzyme involved in oxidative response which has an important role in drug and hormone metabolism in organisms. It is known to be responsible for metabolism of 75% of drugs in humans and is considered a fundamental factor in new drug development.
To activate cytochrome P450, the enzyme must receive an electron by reducing the enzyme. In addition, NADPH (a coenzyme) needs to be present. However, since NADPH is expensive, the use of cytochrome P450 was limited to the laboratory and has not yet been commercialized.
The research team used photosensitizer eosin Y instead of NADPH to develop “Whole Cell Photo-Biocatalysis” in bacteria E. coli. By exposing inexpensive eosin Y to light, cytochrome P450 reaction was catalyzed to produce the expensive metabolic material.
Professor Park said, “This research enabled industrial application of cytochrome P450 enzyme, which was previous limited.” He continued, “This technology will help greatly in producing high value medical products using cytochrome P450 enzyme.”
The research was funded by the National Research Foundation of Korea and KAIST's High Risk High Return Project (HRHRP).
Figure 1: Mimetic Diagram of Electron Transfer from Light to Cytochrome P450 Enzyme via Eosin Y, EY
Figure 2: The back cover of Angewandte Chemie published on 12 January 2015, showing the research results
2015.01.26 View 13500