Kwon Joong Na’s research while affiliated with Seoul National University and other places

Background Spatial transcriptomics (ST) enables the integration of gene expression data with spatial context in tissue samples, providing high-resolution insights without requiring single-cell dissociation. With subcellular resolution, ST offers a precise method for annotating cell types, surpassing traditional approaches based on manual pathologic annotation. Conventional AI models for pathology images typically rely on annotations from pathologists to label cellular compositions in hematoxylin and eosin (H&E)-stained slides. In contrast, our AI model leverages training data generated from subcellular resolution ST, providing a distinct advantage by achieving higher accuracy and objectivity in identifying specific cell types. In this study, we focus on lymphocyte identification in non-small cell lung cancer (NSCLC) pathology images to demonstrate the capabilities of this approach. Methods This model analyzes H&E slide images obtained from surgical specimens of NSCLC patients. The training dataset consisted of ST data (Xenium, 10X genomics) and H&E images from 90 NSCLC samples collected from a single institution. After aligning ST data with H&E images, cell-type masks were generated for lymphocyte labeling. For performance evaluation, 456 patches (256 × 256 pixels, 0.45 µm/pixel) were randomly selected from 30 NSCLC H&E slide images. Two pathologists independently annotated lymphocytes in these patches. Of the 456 patches, 355 with consistent annotations between the two pathologists were selected for model evaluation. The consensus annotations served as the reference standard to assess the model's performance in lymphocyte identification. Results This model achieved an AUROC of 0.8411, indicating high diagnostic accuracy. The optional threshold was determined to be 0.3870, with specificity and sensitivity of 72.76% and 78.92%. Compared to two pathologists’ consensus annotations, the model showed a sensitivity of 81.66%, specificity of 69.70%, and accuracy of 77.29% for lymphocyte identification. Conclusion This AI model trained from subcellular ST provides a robust performance for analyzing the distribution of lymphocytes in NSCLC H&E images. This model offers an innovative approach to analyzing the composition of cells from H&E images, demonstrating its potential contribution to TME research and the development of precision medicine. Citation Format Seo Hye Park, Haenara Shin, Hosub Park, Jaemoon Koh, Kwon Joong Na, Hongyoon Choi. Development and validation of an AI-based model for lymphocyte identification in NSCLC H&E image using spatial transcriptomics [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 2421.

Background Antibody-drug conjugates (ADCs) are targeted therapies that deliver cytotoxic payloads to cells expressing specific antigens. Despite their increasing use in clinical trials for advanced non-small cell lung cancer (NSCLC), many patients experience suboptimal outcomes. While ADC efficacy depends on factors such as antibodies, linkers and payloads, the impact of the tumor microenvironment (TME) remains unclear. This study uses spatial transcriptomics (ST) to investigate the spatial dynamics influencing ADC efficacy within the TME of NSCLC. Methods Visium ST datasets from 36 lung adenocarcinoma (LUAD) tissues were analyzed to identify factors influencing ADC efficacy. Key indices included Target Specificity Score (TSS), which reflects the ratio of target expression in malignant versus normal epithelial cells; Target-Linker Match Score (TLM), which measures overlap between target regions and linker enzyme activity; Therapeutic Index (TI), calculated as TSS × TLM; and Intratumoral Target Expression (ITE), which quantifies average target expression. To evaluate the bystander effect, Moran's I index assessed spatial clustering patterns of target expression, and Minimum Target-High Distance (MTHD) assessed the average of the minimum distance from low to high target expression regions. Fifteen ADCs were selected from NSCLC clinical trial data, including targets, linkers, and overall response rates (ORRs). Correlations between spatial indices and ORRs were analyzed to validate the utility of the indices. Results Of the 36 tissues analyzed, 3 were excluded due to insufficient or excessive tumor regions. Among the six indices evaluated, only MTHD showed a strong negative correlation with the reported ORRs (Spearman's R: -0.715, p = 0.003). This suggests that shorter distances between low and high target expression regions are associated with improved ADC efficacy, highlighting the critical role of spatial homogeneity in the bystander effect. To further evaluate the potential of the indices in identifying ADCs with poor therapeutic efficacy, we compared the indices between ADCs with ORR values below the 10th percentile (ORR Low) and those above (ORR High); only the TSS showed a significant difference (p < 0.05). ROC analysis identified a TSS cut-off value of 4.744, yielding an AUC of 0.962 for discriminating between ORR Low and High groups. Conclusion Spatial indices derived from spatial transcriptomics, particularly those reflecting target specificity and bystander effect, were found to significantly influence treatment response and predict ADC failure, respectively. This in silico modeling approach offers a promising screening method to identify ADCs with the highest potential for therapeutic efficacy. Citation Format Sungwoo Bae, Minki Choi, Hongyoon Choi, Dongjoo Lee, Kwon Joong Na, Daeseung Lee. Spatial transcriptomics-driven in silico modeling of therapeutic effects of antibody-drug conjugates and their association with treatment response in lung adenocarcinoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 6259.

Background Advances in spatial transcriptomics (ST) enable precise gene expression analysis within the tumor microenvironment (TME). VisiumHD and Xenium Prime 5K are cutting-edge platforms offering unique strengths in spatial resolution, sensitivity, and multiplexity. This study evaluates their performance metrics in heterogeneous tumor tissues from the same slide, focusing on technical performance, spatial resolution, and cell type identification. Additionally, post-Xenium VisiumHD datasets were generated and integrated to harness the complementary strengths of both platforms. We aimed to investigate a head-to-head comparison in spatially registered ST to provide guidance for researchers in selecting the appropriate platform for specific cancer research objectives. Methods ST data were analyzed from tumor samples processed with Xenium Prime 5K followed by VisiumHD. Evaluations included transcript capture efficiency, gene detection sensitivity, and signal-to-noise ratios. Spatial resolution was assessed using cell- and transcript-level metrics, gene expression correlations, and spatial mapping accuracy. Cell type identification focused on assignment accuracy and rare population detection. Custom workflows integrated post-Xenium VisiumHD datasets, combining platform strengths for detailed TME analysis. Results Gene expression correlations were highly consistent across platforms. VisiumHD provided superior transcriptome-wide coverage (∼18, 085 genes), making it ideal for exploratory studies, while Xenium offered 2-3x greater per-gene sensitivity for detecting low-abundance transcripts. Xenium Prime 5K excelled in spatial resolution, with precise cell-level annotations, whereas VisiumHD mapped broader spatial patterns via deconvolution. Integration of Xenium cell segmentation with VisiumHD enabled single-cell transcript analysis with comprehensive gene expression data. This approach improved rare cell population detection and overcame limitations of individual platforms. Conclusions VisiumHD is ideal for transcriptome-wide exploratory studies, while Xenium Prime 5K excels in high-resolution, focused cellular analyses. Cross-platform integration bridges gaps between resolution and coverage, enabling robust analysis of complex TMEs. This study provides critical guidance for platform selection and highlights the potential of integrated approaches to advance cancer research through spatial transcriptomics. Citation Format Dongjoo Lee, Yeonjae Jung, Jae Eun Lee, Myunghyun Lim, Daeseung Lee, Kwon Joong Na, Hongyoon Choi. Comparative analysis and integration of image-based and high-resolution sequence-based spatial transcriptomics platforms for tumor microenvironment analysis [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 6276.

Background Spatial biology enables the analysis of tumor microenvironments by revealing cellular relationships within the heterogeneous tumor landscape, which are closely associated with tumor characteristics and immunology. In this study, we utilized high-resolution H&E images and image-based spatial transcriptomics (ST) data from lung cancer to develop a highly accurate deep learning model for mapping and AI-based workflow analyzing spatial relationships in the tumor microenvironment using only H&E images. Methods A total of 164 NSCLC samples with high-resolution ST data (Xenium) were used for model training. Tissue microarrays (TMAs) were prepared, and image and ST data were registered using manual keypoints followed by affine transformation. Refined and major cell types were mapped using label transfer based on ST data aligned with scRNA-seq references. Segmented cell-type maps served as ground truth for training the deep learning model to predict detailed cell-type distributions from H&E images. Model performance was evaluated using the area under the ROC curve (AUROC) for each cell type. External validation was conducted using independent H&E-ST datasets (6 TMA cores and 4 whole-slide images). The trained models were integrated with a customized StarDist-based nucleus segmentation tool for H&E images. Results In the external validation datasets, the model achieved mean AUROC values of 0.96, 0.92, 0.95, 0.94, 0.95, and 0.94 for epithelial cells, myeloid cells, fibroblasts, T-cells, endothelial cells, and B-cells, respectively. For refined immune cell types, including CD4+ T-cells, CD8+ T-cells, dendritic cells, and NK cells, the AUROC values were 0.94, 0.97, 0.94, and 0.97, respectively. The segmented cell-type maps were integrated with nucleus segmentation to generate spatial data formats comprising cell types, enrichment scores, and spatial coordinates. This integrated deep learning model provided an H&E-based AI tool for spatial biology, enabling the analysis of spatial relationships in the tumor microenvironment, such as cell type density in tumor epithelial niches. Conclusion Large-scale image-based ST data with precisely registered H&E images can achieve highly accurate cell-type definitions, even for refined immune cell types like CD4+ and CD8+ T-cells and dendritic cells. This integrated workflow offers a powerful tool for analyzing spatial features using only H&E images in tumor biology. Citation Format Haenara Shin, Dongjoo Lee, Yooeun Kim, Daeseung Lee, Kwon Joong Na, Hongyoon Choi. High-resolution AI-based spatial biology tool for lung cancer trained by image-based spatial transcriptomics data to analyze tumor microenvironmnet [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 2443.

Unraveling the spatial configuration of the tumor microenvironment (TME) is crucial for elucidating tumor-immune interactions based on immuno-oncology. We present STopover, a novel approach utilizing spatially resolved transcriptomics (SRT) data and topological analysis to investigate the TME. By gradually lowering the feature threshold, connected components (CCs) are extracted based on spatial distance and persistence, with Jaccard indices quantifying their spatial overlap, and transcriptomic profiles are permutated to assess statistical significance. Applied to lung and breast cancer SRT, STopover revealed immune and stromal cell infiltration patterns, predicted key cell–cell communication, and identified relevant regions, shedding light on cancer pathophysiology (URL: https://github.com/bsungwoo/STopover). Supplementary Information The online version contains supplementary material available at 10.1186/s13073-025-01457-1.

Background: Digital high-resolution H&E images provide valuable insights into tumor heterogeneity within the tumor microenvironment (TME). However, the ability to perform detailed cell typing solely based on H&E images remains limited. Recent advances in high-resolution spatial transcriptomics (ST) enable precise characterization of cell types and their spatial relationships within the TME, addressing challenges in understanding tumor heterogeneity. In this study, we developed AI models to predict cell types in breast cancer—including subtypes of lymphocytes that are challenging to differentiate visually—using large-scale image-based ST data aligned with H&E images. Methods: We established a breast cancer image-based ST database comprising 190 samples from 113 breast cancer patients, obtained from surgically resected primary tumors. Image-based ST data were generated using the Xenium platform with a 500-gene panel and matched with high-resolution H&E images. Cell type maps were constructed using reference single-cell RNA-seq data and transferred to ST data. The ST data were spatially registered to the corresponding H&E images, and cell type masks were generated at matching resolutions. AI models were trained to segment cell type masks, including epithelial cells, cancer cells, myeloid cells, fibroblasts, endothelial cells, T cells, and B cells. Additionally, refined cell type predictions were performed for dendritic cells, NK cells, CD4+ T cells, and CD8+ T cells. Model performance was validated using external whole-slide image datasets containing ST data from four independent cohorts. Results: Model performance was assessed using the area under the receiver operating characteristic (AUROC) curve for each cell type mask. Internal validation on four samples yielded AUROC values ranging from 0.90 to 0.95. External validation across independent whole-slide datasets demonstrated AUROC values between 0.88 and 0.97 for all cell type masks. The AI model successfully mapped TME cell types with high resolution, using only H&E images. Conclusion: By employing a self-supervised approach that integrates high-resolution H&E images with image-based ST data, we developed AI-driven tools for TME analysis. These models enable accurate identification of detailed cell types and their spatial relationships within the TME. This approach facilitates large-scale analysis of breast cancer TME and holds potential for advancing our understanding of tumor biology and therapeutic strategies. Citation Format: Haenara Shin, Dongjoo Lee, Yooeun Kim, Daeseung Lee, Kwon Joong Na, Chihwan David Cha, Hosub Park, Hongyoon Choi. A self-supervised AI model leveraging spatial omics for analyzing tumor microenvironment heterogeneity in breast cancer only with H&E [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Functional and Genomic Precision Medicine in Cancer: Different Perspectives, Common Goals; 2025 Mar 11-13; Boston, MA. Philadelphia (PA): AACR; Cancer Res 2025;85(5 Suppl):Abstract nr B038.

Objectives Recently, sarcopenia has been linked to unfavorable outcomes in various surgical procedures, including lung cancer surgery. This study aimed to investigate the impact of respiratory sarcopenia (RS) on postoperative and long-term outcomes in elderly patients undergoing lung cancer surgery. Methods This retrospective study included patients aged 70 years and older who underwent lobectomy with curative intent for lung cancer between 2017 and 2019. RS was defined as having values below the median for both the L3 skeletal muscle index, measured from preoperative PET-CT images, and peak expiratory flow (PEF). An inverse probability of treatment weighting (IPTW) approach was applied to balance covariates between the RS and non-RS groups. Baseline characteristics and postoperative outcomes were compared between groups using t-tests and chi-square tests. Kaplan–Meier curves and log-rank tests were used to compare overall and recurrence-free survival. Multivariable logistic regression analysis incorporating IPTW weights was performed to assess the impact of RS on respiratory complications. Results A total of 509 patients were included, of whom 123 (24.2%) had RS. After IPTW adjustment, baseline characteristics, including pulmonary function, were similar between the RS and non-RS groups. All patients underwent lobectomy, with 78.8% of the RS group and 80.9% of the non-RS group undergoing minimally invasive surgery. The RS group had a significantly higher rate of respiratory complications compared to the non-RS group (14.5% vs. 7.7%, p = 0.041). Multivariable logistic regression analysis showed that male sex (odds ratio = 15.2, p < 0.01) and lower DLCO (odds ratio = 0.96, p < 0.01) were significantly associated with respiratory complications, whereas RS did not show a significant association (p = 0.05). No significant differences were found in overall survival (p = 0.11) or recurrence-free survival (p = 0.51) between the groups. Conclusions In this study, RS had a limited impact on both postoperative and long-term outcomes in elderly patients undergoing lung cancer surgery. These findings suggest that other factors, such as DLCO and male sex, may play a more prominent role in predicting respiratory complications.

Background: Posterior reversible encephalopathy syndrome (PRES) is a rare complication of lung transplantation with poorly understood risk factors and clinical characteristics. This study aimed to examine the occurrence, risk factors, and clinical data of patients who developed PRES following lung transplantation.Methods: A retrospective analysis was conducted on 147 patients who underwent lung transplantation between February 2013 and December 2023. The patients were diagnosed with PRES based on the clinical symptoms and radiological findings. We compared the baseline characteristics and clinical information, including primary lung diseases and immunosuppressive therapy related to lung transplantation operations, between the PRES and non-PRES groups.Results: PRES manifested in 7.5% (n=11) of the patients who underwent lung transplantation, with a median onset of 15 days after operation. Seizures were identified as the predominant clinical manifestation (81.8%, n=9) in the group diagnosed with PRES. All patients diagnosed with PRES recovered fully. Patients with PRES were significantly associated with connective tissue disease-associated interstitial lung disease (45.5% vs. 18.4%, P=0.019, odds ratio=9.808; 95% CI, 1.064–90.386; P=0.044). Nonetheless, no significant variance was observed in the type of immunotherapy, such as the use of calcineurin inhibitors, blood pressure, or acute renal failure subsequent to lung transplantation.Conclusions: PRES typically manifests shortly after lung transplantation, with seizures being the predominant initial symptom. The presence of preexisting connective tissue disease as the primary lung disease represents a significant risk factor for PRES following lung transplantation.

OBJECTIVES This study compared the mid-term patency of expanded polytetrafluoroethylene grafts without rings versus that of bovine pericardial conduits used for superior vena cava reconstruction for various thoracic diseases. METHODS Among 80 patients who underwent superior vena cava resection and reconstruction between 2009 and 2023 at our institution, 31 patients who received polytetrafluoroethylene grafts without rings (Polytetrafluoroethylene group) and 28 patients who received bovine pericardial conduits (Bovine group) were enrolled. Median follow-up durations were 19.5 and 64.6 months in the Polytetrafluoroethylene and Bovine groups, respectively. Primary outcome was midterm graft patency rate, and secondary outcomes were early and midterm clinical outcomes, including all-cause mortality and superior vena cava reintervention. RESULTS Operative mortality was 1.7%. Cumulative incidence of all-cause mortality was not significantly different between the groups. Graft occlusion was detected in 22 patients. Cumulative incidence of graft occlusion was 24.2%, 36.4%, 42.4%, 48.5% and 60.6% at 1 month, 3 months, 6 months, 1 year and 2 years, respectively, in the Bovine group, whereas no graft occlusion was observed in the Polytetrafluoroethylene group (P = 0.007). Although the incidence of graft occlusion was higher in the Bovine group, cumulative incidence of reintervention was not significantly different between the groups (0.0% vs 3.0% in Polytetrafluoroethylene vs Bovine groups at 1 year, P = 0.406). Multivariate analysis demonstrated that bovine pericardial conduit (polytetrafluoroethylene graft as reference) and left brachiocephalic vein reconstruction (right brachiocephalic vein reconstruction as reference) were significant risk factors for graft occlusion. CONCLUSIONS In superior vena cava reconstruction, polytetrafluoroethylene grafts without rings were superior to bovine pericardial conduits in terms of midterm graft patency.