Park, Sunho

Abstract Background: Immune checkpoint inhibitors (ICIs) targeting PD-1 or PD-L1 have dramatically improved survival of patients with gastric cancer (GC). However, recent clinical trials, including KEYNOTE-059, KEYNOTE-061, and ATTRACTION-2, show that the objective response rate (ORR) for ICIs is still less than 20% in patients with advanced GC. Therefore, predicting who will respond to ICIs in advance is essential to make better treatment plans for GC patients. There are potential biomarkers to predict ICI response, but most of them can only be applied to a limited population of patients. For example, Kim ST., et al., (PMID: 30013197) show that patients with MSI-high or EBV-positive responded well to ICIs, but these subtypes account for less than 30 percent of GC patients (PMID:25079317). In this study, we propose a prediction model of ICI response using the 32 GC-specific genes that we identified from mutation data (PMID: 26635139) and showed their close association with survival and ICI response in GC (PMID: 35140202; PMID: 36508166). In addition, since the number of available GC-ICI samples is small, we design our prediction model in Bayesian and ensemble learning frameworks which are robust to small sample size problems. Methods: We collected mRNA sequencing data from GC-ICI treated patients from our institutions and publications (PMID: 30013197; PMID: 35254400; PMID: 33846173). We used 108 samples previously included in our previous work (PMID: 36508166) as the training data and 27 samples (13 responders and 14 non-responders) from (PMID: 3384617) as the test data. We built and trained a classifier to separate responders from non-responders using the expressions of the 32 genes. We proposed to use a Deep Gaussian process (GP) model, a multi-layered extension of a GP that is itself a Bayesian method. We optimized the parameters of each GP model in variational inference. To implement ensemble learning, we trained multiple DGP models with different initializations of the model parameters and data splitting (training/validation sets). For each training run, the best model was selected based on the prediction performance on the validation set, measured by AUC. The final prediction for a test sample was made by combining the prediction results of all the DGP models by taking the average. Results: We implemented the DGP-based prediction model using PyTorch in Python, and the codes will be publicly available on the GitHub repository (https://github.com/hwanglab/GC_ICI_response_prediction). We achieved a prediction performance of 0.813 AUC on the test data (sensitivity = 0.769 and specificity = 0.714). Conclusion: We propose the prediction model of ICI response, which can handle small sample size problems in the framework of Bayesian and ensemble learning. The experimental results using the GC-ICI data demonstrate that the 32 genes can be used as predictive markers for ICI response in GC. Citation Format: Sunho Park, Minji Kim, Changjin Hong, Tae Hyun Hwang. A deep Gaussian process (DGP)-based prediction model for immunecheckpoint inhibitor (ICI) response in gastric cancer (GC) [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 2316.

Abstract Background: Microsatellite instability-high (MSI-H) tumors respond well to immune checkpoint blockade and oftentimes not to chemotherapeutics. Despite value of determining MSI status to guide therapy, many patients’ samples are not untested. Deep learning models may be able to predict MSI status by analyzing hematoxylin and eosin (H&E) stained whole-slide images (WSIs). Validation of this tool will provide another tool to increase MSI testing. However, to be deployed into clinical routine care, prediction models need be externally validated in large and diverse patient cohorts. In addition, these models should provide uncertainty of prediction to help clinicians make informed decisions. Results: In this study, we introduce a novel Bayesian prediction model using whole-slide images (WSIs) and deep Gaussian processes (DGPs) in a weakly supervised learning context. Extensively tested on large datasets (n=3,146) from multiple centers, our model not only demonstrated superior performance but also showed that incorporating uncertainty into predictions significantly improves MSI prediction accuracy, achieving AUROC of >0.91 in colon cancer and >0.94 in gastric cancer. Additionally, an in-depth analysis within the gastric cancer cohort indicated a significant correlation between predicted MSI-H regions and immunotherapy response. Patients who responded to immune checkpoint inhibitors had a higher proportion of MSI-H regions (P-value < 0.05). This correlation was also observed in patients categorized as MSS but responsive to ICIs, highlighting the potential of our model to identify MSI-H patterns not detected through standard laboratory tests. This advancement could lead to more precise patient selection for ICI therapy, enhancing personalized treatment strategies in gastrointestinal cancer. Citation Format: Sunho Park, Minj Kim, Ji-Youn Sung, Jeong Hwan Park, Sung Hak Lee, Sam C. Wang, Jaeho Cheong, Tae Hyun Hwang. Explainable AI model incorporating uncertainty estimation for enhanced MSI-H prediction and immunotherapy response in gastrointestinal cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 7385.

Abstract Traditional cancer research methodologies predominantly focus on 2D tumor modeling, including spatial profiling and AI-driven analysis at the 2D level. To advance beyond these limitations, we implemented Holotomography (HT), a 3D, label-free optical imaging method. This technique, leveraging refractive index contrasts, allows for enhanced visualization of tumor tissues, organoids, and cells in three dimensions. Using serial H&E sections and unstained slides, we created detailed 3D tumor models in gastric cancer, providing insights into tumor, immune, and stromal composition, as well as cell-cell interactions and extracellular matrix characteristics. AI algorithms further enhanced our 3D tumor modeling, offering novel perspectives on the initiation and progression of tumor, immune, and stromal components at the subcellular level. Analyzing gastric tumors treated with immune checkpoint inhibitors before, during, and after treatment revealed dynamic changes in spatial composition over time, offering unique insights into the evolution of the 3D tumor ecosystem. Additionally, by integrating spatial transcriptomic and in situ molecular imaging data, we developed a novel Multimodal 3D Tumor Model. This model not only uncovers subcellular-level molecular characteristics but also elucidates cellular interactions during treatment, suggesting potential biomarkers and therapeutic approaches. Lastly, by constructing 3D cellular interaction networks incorporating true 3D interactions, we provided an in-depth view of cellular communication within the tumor microenvironment. This pioneering approach, representing the first Multimodal 3D Tumor Model at the subcellular level, offers novel insights into gastric cancer's spatial evolution, potentially revolutionizing our understanding and treatment of gastric cancer. Citation Format: Yoon Ho Choi, Minji Kim, Katya Redekop, Inyeop Jang, Jean R. Clemenceau, Isabel Barnfather, Sumanth Reddy Nakkireddy, Changjin Hong, Sunho Park, Ji-Youn Sung, Jeong Hwan Park, Shitara Kohei, Aoki Yu, Sung Hak Lee, Sam Wang, Jaeho Cheong, Tae Hyun Hwang. AI-driven multimodal 3D tumor modeling with spatial molecular insights for predicting ICI response in gastric cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 7400.

Abstract Background: The Tumor Immune microenvironment (TME) plays an important role in tumor development, progression, and treatment response. Recent studies show that spatial heterogeneity of Tumor-Infiltrating Lymphocytes (TILs), consisting of lymphocytic cell populations, is an important histopathologic feature and a powerful prognostic and predictive marker correlating with cancer progression and treatment response. Detecting and quantifying TIL density in Whole Slide Images (WSIs) is, however, labor-intensive, and it is difficult to assess differences in the composition of immune cells. Therefore, it remains a challenge to establish the analysis of TILs in the complex TME. We present an automated spatial image analysis of TILs on gastric cancer to address cancer immunotherapy. Method: 1,031WSIs of hematoxylin and eosin-stained (H&E) formalin-fixed paraffin-embedded (FFPE) sections from four patient cohorts: TCGA-STAD (329), Yonsei-1 (620 patients), Yonsei-2 (45), and St. Mary (37), were analyzed. Two treatment groups in response were classified as CR + PR, while another two treatment groups in non-response were classified as PD + SD. H&E stained WSIs were normalized using the Reinhard normalization method at 20X magnification for reducing the color variability of the stained tissue images. We used U-Net architecture to predict the normalized H&E stained WSIs. Three U-Net models were trained for predicting tumor, TILs, and stroma, using Adam optimizer, 1e-4 learning rate, a binary loss function, L2 regularization, and 50 epochs. The normalized H&E stained WSIs were predicted using the trained U-Net models. Spatial analysis of TILs was conducted on the predicted WSIs. Regional quantification of TILs was performed by two immune phenotype groups: immune-immersed (TILs enriched in tumor cells) and immune-excluded (TILs not enriched in tumor cells but in stromal cells). Kaplan-Meier with a log-rank test was used to evaluate the survival differences between the groups. Results: In TCGA-STAD, the proportion of immune-immersed composition is significantly different from the immune-excluded (p-value 0.05). In Yonsei-1, the immune-excluded composition is significantly different among the three risk groups (p-value 3e-05). In the Yonsei-2 and St. Mary cohorts, including anti-PD1 treated gastric cancer patients, the immune-immersed composition is significantly different between two response groups (p-value 0.007) and represented the diagnostic accuracy (AUROC: 0.78) on CR + PR groups. Conclusion: These findings suggest that the proposed TIL quantification in the complex TME is strongly associated with patient prognosis and anti-PD1 treatment response in gastric cancer. Citation Format: Sanghoon Lee, Sunho Park, Jae-Ho Cheong, Sam C. Wang, Matthew R. Porembka, In-Ho Kim, Sung Hak Lee, Tae Hyun Hwang. Spatial image analysis of tumor-infiltrating lymphocytes on gastric cancer to predict anti-PD1 inhibitor response [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 1360.

Abstract Background: While immune checkpoint inhibitors (ICI) extend survival for patients suffering from a variety of cancers, most gastric cancer (GC) patients do not respond to ICI. There is a critical need to identify novel biomarkers that guide the use of ICI or strategies to improve outcomes for GC patients. We aim to gain a deeper understanding of the GC tumor microenvironment (TME) to refine the use of ICI using spatial transcriptomes from ICI-treated gastric tumors. Method: We collected 12 pre-ICI treatment gastric tumor samples from five responders and seven nonresponders and performed spatial profiling to investigate spatial heterogeneity, functions, and phenotypes of tumor and neighboring stromal and immune cells to identify features associated with ICI response. We used pan-cytokeratin (pan-CK) to mark epithelial cells, CD45 for immune cells, and smooth muscle actin (SMA) for fibroblasts. We generated gene expression profiles of each cell type within regions of interest (ROI) with a NanoString GeoMx Cancer Transcriptome Atlas panel, comprised of 1,800 genes. A total of 58 ROIs were analyzed from responders and 62 ROIs from nonresponders. Results: We measured the spatial heterogeneity of each cell type in ROIs within and across tumors. We found the tumor, immune, and fibroblasts from nonresponders are transcriptionally more heterogeneous, compared to those from responders. This might indicate a higher degree of intratumor transcriptional heterogeneity is linked to ICI resistance. We found that nonresponse was associated with higher expression of VEGFA and VEGFB in tumor cells and higher ACTA2 in fibroblasts, not in other cell types. These findings are intriguing when considering our other finding that increased endothelial cell and fibroblast proportions in the TME are associated with ICI nonresponse. We also found that the TME of nonresponders had fewer M1 and more M2 macrophages, and there were more T cells that overexpressed gene programs consistent with dysfunction and exhaustion. These findings indicate that ICI-resistant TMEs are densely surrounded by immune-suppressive cellular components which might reduce ICI efficacy and result in ICI resistance. Finally, we performed cell-cell interaction and identified ligand-receptor interactions that are stronger in nonresponders, including CXCL12-CXCR4, CD74-MIF, and LGALS9-LRP1 ligand-receptor interactions. The predicted ligand-receptor interactions are present in cell-type-specific manners, indicating the functional role of each cell type in facilitating ICI response. Conclusion: We investigated how tumor and neighboring stromal and immune cells interact within GC TME to mediate ICI response, which demonstrate that spatial analysis could be used to identify novel biomarkers that predict response to ICI, and new therapies to enhance ICI efficacy in GC patients. Citation Format: Sunho Park, Changjing Hong, Jae-Ho Cheong, Sam C. Wang, Matthew R. Porembka, In-Ho Kim, Sung Hak Lee, Tae Hyun Hwang. Spatial architecture and cellular interactions of tumor immune microenvironment to discover biomarkers and predict immune checkpoint inhibitor response in gastric cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 2262.

Abstract Background: Microsatellite Instability (MSI) is a prognostic and predictive biomarker which can guide treatments including immunotherapy for patients with gastrointestinal cancer. Universal MSI testing would benefit to patients to receive a better therapy, but many patients remain untested. To deliver broadly accessible MSI testing, deep learning models have demonstrated their feasibility for MSI prediction using H&E-stained whole-slide images (WSIs). However, to be deployed into clinical routine care, prediction models should be validated in datasets from multicenter and multiethnic groups. In addition, these models should provide uncertainty of prediction to help clinicians to make informed decisions. Method: We develop a MSI prediction model using WSIs based on Deep Gaussian process (DGP), a Bayesian model being able to model the uncertainty in prediction. We implement a DGP model in the transfer learning, where a WPI is decomposed into multiple non-overlap image patches which are converted to feature vectors by a pretrained convolutional neural network (CNN). Then, the DGP model makes a prediction for an image by averaging score function values at all the feature vectors in the image in weakly supervised learning (WSL). To test our method, we collected H&E stained colorectal/stomach cancer slides (n=1,619) from multi-institutions and multiethnic groups, including TCGA Colorectal (CRC; n=351) and Stomach Adenocarcinoma (STAD; n=174) and six datasets from tertiary hospitals in Korea (i.e., Yonsei-1 (n=174), Yonsei-1-remade (n=146), Yonsei-2 (n=95), St. Mary-1 (n=48), St. Mary-2 (n=50), and Yonsei-CLASSIC trial (n=581)). We also compare our method to multiple deep learning methods, including Densenet, Resnet, Shufflenet, Googlenet, Squeezenet and two WSL models, MIL (Campanell et al., 2019) and CLAM (Lu et al., 2021). Result: In each cancer type, we train each model using TCGA dataset as discovery cohort and tested on the rest of datasets as validation cohort. We use the area under of roc curve (AUC) for the evaluation metric. For the CRC cohorts, our DGP method achieves overall best performance with 0.81 (Yonsei-1), 0.82 (Yonsei-1-remade), 0.89 (Yonsei-2), 0.85 (St. Mary-1) 0.75 (St. Mary-2) AUCs compared to the state-of-the art methods. For the STAD cohorts, our method also achieves overall best performance with 0.74 AUC (Yonsei-classic). In addition, incorporating the uncertainty in prediction measured by our method improves the model’s performance to predict MSI. Specifically, removing cases with prediction results with high uncertainty, which could lead false positive prediction, significantly improves MSI prediction performance. Taken together, these results demonstrate the robustness and generalisability of the DGP model for MSI prediction across multicenter and multiethnic datasets. Citation Format: Sunho Park, Hongming Xu, Sung Hak Lee, Jeonghyun Kang, Tae Hyun Hwang. Deep Gaussian process with uncertainty estimation Improves microsatellite instability prediction based on whole slide image: A retrospective multicenter and multiethnic cohort study [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 5012.