This study is a collaborative effort among four research institutions: Prof. Hyongbum (Henry) Kim’s lab at Yonsei University College of Medicine, Prof. Jung-Hoon Kim’s lab at Seoul National University College of Medicine, Prof. David Liu’s lab at the Broad Institute, and Prof. Young Wook Jun’s lab at UCSF.

Prof. Hyongbum Kim’s lab at Yonsei University conducts comprehensive research on next-generation gene-editing technologies, focusing on the following areas:

This study aims to evaluate the therapeutic efficacy and potential side effects of the latest gene-editing systems—PE7 (Prime Editor 7) and CAST (CRISPR-Associated Transposase)—in mouse models of various intractable genetic diseases, including hemophilia, X-linked retinoschisis, Leber congenital amaurosis, autism spectrum disorder, and congenital hearing loss.

The ultimate goal is to establish curative treatments for genetic disorders affecting the eye, liver, ear, and brain. While current clinical approaches primarily focus on symptom management, this research seeks to directly correct disease-causing mutations using gene-editing tools and to evaluate their therapeutic outcomes and safety profiles. The results are expected to provide critical data for enabling clinical translation of gene-editing therapies.

Intractable genetic disorders often arise from a range of germline mutations. Because curative treatment requires direct mutation correction, most current clinical approaches remain symptomatic.
Gene-editing technologies, which have evolved since their discovery in 2012, can precisely introduce, correct, or insert large genetic materials at targeted DNA sequences. Inspired by these possibilities, our team initiated this preclinical study using the latest platforms—PE7 and CAST—to explore therapeutic applications across multiple genetic diseases.

If successful, this study will mark a significant milestone in applying gene-editing therapy to patients with rare and intractable genetic disorders. It is expected to enable a shift from symptomatic management to precise, mutation-correcting treatments, offering hope of true cures to patients and families affected by these conditions.

One of the key outcomes involves identifying mutations in the ATM gene that increase cancer susceptibility using prime editing (PE).
By introducing 27,513 mutations into the ATM gene via the PE system, the research team evaluated their functional impact through high-throughput screening. Leveraging this dataset, a deep learning model was developed to predict the functional consequences of untested variants.

This work represents a major advancement in rapidly and comprehensively assessing Variants of Unknown Significance (VUS), accelerating the ability to evaluate the relationship between genetic alterations and loss of gene function. The results are expected to provide important clinical guidelines for assessing cancer risk and guiding personalized anticancer drug selection.

Functional assessment of all ATM SNVs using prime editing and deep learning

Kwang Seob Lee et al, (2025) Cell

The goal of this study is to realize precision medicine in breast cancer by integrating spatial omics with multimodal artificial intelligence technologies. By analyzing tumor heterogeneity within the spatial context of the tumor microenvironment, we aim to develop patient-specific biomarkers for diagnosis, treatment, and recurrence prediction.

Contrary to common belief, a tumor is not composed of identical cancer cells. Within it are both “responsive” cells that react well to therapy and “resistant” cells that persist despite treatment. Our research employs AI navigation to map tumor tissues and locate regions where these resistant cells hide. Using laser-based precision isolation technology, we extract and analyze only those localized cells to uncover their molecular features. This approach enables us to reveal the fundamental causes of cancer recurrence and to pioneer personalized treatment strategies tailored to each patient—a step toward a new era of customized breast cancer therapy.

The greatest challenge in cancer treatment lies in tumor heterogeneity. Even within a single tumor, drug-sensitive and drug-resistant cancer cells coexist. A small population of resistant cells can survive chemotherapy and trigger recurrence. These cells often cluster in specific spatial niches within the tumor microenvironment, making them difficult to detect and characterize using conventional techniques.

To overcome this limitation, our research integrates AI-based risk prediction with spatial omics to precisely identify and analyze the molecular properties of resistant cancer cells. Furthermore, to translate our findings into clinical application, we utilize organoid models and liquid biopsy technologies to explore novel therapeutic approaches and develop easy-to-perform blood-based recurrence detection methods, supporting the implementation of truly practical precision oncology.

In the long run, this research will accelerate the transition of cancer therapy from the conventional “standard-of-care” paradigm to an era of “ultra-precise personalized medicine.” This shift will have profound industrial implications, enhancing success rates in drug development by revealing the mechanisms of drug resistance and facilitating the discovery of new therapeutic targets.

Ultimately, by being the first in the world to clinically apply Korea’s combined strengths in AI and biotechnology through an innovative medical platform, this project is expected to establish Korea as a global leader in precision medicine while generating substantial socioeconomic value.

Our team has been actively investigating the interplay between tumor heterogeneity and the immune microenvironment. A major achievement was demonstrating that PD-L1 expression levels are a key determinant of immune checkpoint inhibitor response in breast cancer patients. The findings provide critical evidence for developing personalized immunotherapy strategies, particularly for triple-negative breast cancer. We also revealed that excessive PD-L1 expression suppresses T-cell infiltration and activation, offering insights into the functional dynamics of the tumor-immune interface.

We have generated an extensive dataset exceeding 140–200% of initial goals, now encompassing approximately 600 cases of matched pathology images and genomic data. Based on this dataset, our recurrence prediction model achieved a performance improvement of over 8% compared to existing benchmarks (C-index +0.054) and successfully produced clinically interpretable RNA maps directly from pathology images.

The most striking discovery was that the conventional H&E-stained images—used for decades in pathology—harbor vast, previously untapped molecular information. Deep AI analysis of these images revealed that molecular-level data, such as the expression of specific genes, can be inferred directly from tissue morphology. This discovery highlights a profound link between structural and molecular characteristics, suggesting a transformative shift in how cancer diagnostics and pathology will be conducted in the future.

The Center for Emerging Viral Hemorrhagic Fever Research focuses on combating infections caused by novel and re-emerging viral pathogens with high fatality rates and epidemic potential, such as Severe Fever with Thrombocytopenia Syndrome Virus (SFTSV) and Hantaan virus. These viruses are particularly dangerous for the elderly and immunocompromised, as they can cause life-threatening hemorrhagic fevers.

Through the collaboration of Korean and U.S. researchers, clinical specimens—including patient blood, bone marrow cells, and cerebrospinal fluid—are collected and analyzed to elucidate the mechanisms of disease progression and immune response. The research aims to uncover the biological principles of viral pathogenesis, develop and stabilize new vaccine candidates, discover broad-spectrum antiviral agents, and validate their efficacy through animal models.

Ultimately, the project seeks to establish predictive biomarkers for disease progression, identify antiviral and anti-inflammatory drug targets, and develop multivalent vaccine platforms capable of protecting against multiple viral species. These efforts aim to strengthen public health preparedness and medical security at both national and global levels.

As demonstrated during the COVID-19 pandemic, sudden outbreaks and the rapid spread of novel viral pathogens can cause devastating consequences worldwide. Such outbreaks have occurred repeatedly throughout history, and predicting the next causative virus remains extremely challenging. Therefore, there is an urgent need to build a scientific foundation for preventive and therapeutic strategies targeting highly fatal RNA viruses that pose global health threats.

This project brings together leading U.S. and Korean experts who have long studied hemorrhagic fever viruses to collaboratively develop universal preventive and therapeutic technology platforms. These platforms are designed to enhance preparedness and effectively respond to future infectious disease crises.