Background:This study aimed to determine the prevalence of human epidermal growth factor receptor 2 (HER2)-ultralow breast cancer among cases initially classified as HER2 immunohistochemistry (IHC) 0 and assess interobserver variability in interpreting low-level HER2 expression. Methods: In this multicenter retrospective study, all invasive breast cancer cases diagnosed between January and December 2022 across 10 Korean institutions were retrieved. Institutional pathologists reexamined HER2 IHC slides originally reported as IHC 0 according to the 2018 American Society of Clinical Oncology/College of American Pathologists guidelines and reclassified them as HER2-null (0), HER2-ultralow (0+), or HER2-low (1+). Slides from 10% of HER2-null and HER2-ultralow cases were digitized for central review and independently assessed by two pathologists, with discrepancies resolved by consensus. Results: Among 8,026 cases, 2,836 cases (35.5%) were initially reported as IHC 0. Upon re-review, 1,673 (59.0%), 1,139 (40.2%), and 24 (0.8%) cases were reclassified as HER2-null, HER2-ultralow, and HER2-low, respectively. The prevalence of HER2-ultralow breast cancer varied considerably across institutions (23.7%-78.1%). Central review of 268 digitized cases showed concordance in 193 cases (72.0%). Among the 75 discordant cases, 54 tumors (72.0%) were upgraded from HER2-null to HER2-ultralow, and 18 (24.0%) tumors were upgraded from HER2-ultralow to HER2-low. Furthermore, two tumors (2.7%) were downgraded from HER2-ultralow to HER2-null. Conclusions: Approximately 40% of cases initially categorized as IHC 0 were reclassified as HER2-ultralow. The substantial inter-institutional variability observed in interpreting low-level HER2 expression highlights the need for standardized training and quality assurance to ensure accurate identification of patients eligible for HER2-targeted antibody-drug conjugates.

Identifying the cell of origin that harbors an initial driver mutation is key to understanding tumor evolution and for the development of new treatments. For isocitrate dehydrogenase (IDH)-mutant gliomas, the most common malignant primary brain tumor in young adults, the cell of origin is currently poorly understood. We conducted deep sequencing on 142 tissues from 70 individuals comprising tumors, peritumoral cortex or subventricular zones, and blood. Low-level IDH mutations were found in the peritumoral cortex in 37.9% (11 of 29) of patients. Integrating cell-type-specific mutation analysis, the direction of clonal evolution, spatial transcriptomics from patient brains, and a cancer mouse model arising from mutant oligodendrocyte progenitor cell, we determined that glial progenitor cells harboring an initial IDH mutation were responsible for the development of IDH-mutant gliomas.

To ensure high-quality bioresources and standardize biobanks, there is an urgent need to develop and disseminate educational training programs in accordance with ISO 20387, which was developed in 2018. The standardization of biobank education programs is also required to train biobank experts. The subdivision of categories and levels of education is necessary for jobs such as operations manager (bank president), quality manager, practitioner, and administrator. Essential training includes programs tailored for beginner, intermediate, and advanced practitioners, along with customized training for operations managers. We reviewed and studied ways to develop an appropriate range of education and training opportunities for standard biobanking education and the training of experts based on KS J ISO 20387. We propose more systematic and professional biobanking training programs in accordance with ISO 20387, in addition to the certification programs of the National Biobank and the Korean Laboratory Accreditation System. We suggest various training programs appropriate to a student’s affiliation or work, such as university biobanking specialized education, short-term job training at unit biobanks, biobank research institute symposiums by the Korean Society of Pathologists, and education programs for biobankers and researchers. Through these various education programs, we expect that Korean biobanks will satisfy global standards, meet the needs of users and researchers, and contribute to the advancement of science.

The efficacy of treating colorectal cancer (CRC) liver metastases is hindered by significant tumor heterogeneity, both within and between tumors. This diversity stems from varied cell populations and complex interactions in the tumor microenvironment, necessitating a deeper understanding of spatial organization and cellular communication patterns. We examined 23 formalin-fixed paraffin-embedded tissue samples from 11 patients, comprising primary lesions and matched liver metastasis (6 with synchronous and 5 with metachronous metastases) using NanoString 6K CosMx Spatial Molecular Imaging (SMI). Pathologists selected optimal fields of view at tumor invasive fronts. Image processing, cell segmentation, and feature extraction utilized an in-house CosMx SMI pipeline with the Cellpose algorithm. Following quality control, the remaining cells underwent unsupervised clustering with differential gene expression analysis and cell-to-cell interaction analyses performed using the Seurat R package. Analysis of 105, 624 high-quality cells revealed 23 distinct clusters, categorized as ten epithelial tumors, three cancer-associated fibroblast (CAF), two endothelial cells, five immune cells (T cell, Monocyte, Macrophage, SPP1+ Macrophage, and Plasma cell), alveolar cells, hepatocytes, and low complexity cell populations. We identified significant differences in gene expression between colon and liver CAFs, with liver CAFs displaying upregulation of 92 genes associated with epithelial-mesenchymal transition (EMT). Enhanced communication between CAFs, T cells, and macrophages was observed in liver metastases, with specific ligand-receptor pairs (SPP1-ITGA4/ITGB1, LGALS9-P4HB, LGALS9-CD44, CXCL12-CXCR4) enriched in the metastatic microenvironment. Our findings demonstrate distinct spatial heterogeneity and intercellular communication networks in CRC liver metastases. The identification of liver-specific CAF gene expression profiles and unique ligand-receptor interactions provides new insights into metastatic mechanisms and potential therapeutic targets. Jongwon Lee, Yeseul Kim, Hyo Seon Ryu, Jongmin Sim, Chungyeul Kim, Hyun Woo Kwon, Hyun Je Kim, Jong Min Park, Ah-Reum Lim, Jung Sun Kim, Hwa Jung Sung, Xingyi Guo, Jungmin Choi, Jungyoon Choi. Mechanisms of metastasis revealed through single-cell spatial transcriptome analysis in patients with liver metastatic colorectal cancer [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 5289.

Discovering the cell-of-origin harboring the initial driver mutation provides a fundamental basis for understanding tumor evolution and development of new treatments. For isocitrate dehydrogenase (IDH) mutant gliomas, the most common malignant primary brain tumors in adults under 50, the cell-of-origin remains poorly understood. Here, using patient brain tissues and genome-edited mice, we identified glial progenitor cells (GPCs), including oligodendrocyte progenitor cells (OPCs), as the glioma-originating cell type harboring the IDH mutation as the initial driver mutation. We conducted comprehensive deep sequencing, including droplet digital PCR and deep panel and amplicon sequencing to 128 tissues from 62 patients (29 IDH-mutant gliomas and 33 IDH-negative controls) comprising tumors, normal cortex or normal subventricular zone (SVZ), and blood. Surprisingly, low-level IDH mutation was found in the normal cortex away from the tumor in 38.5% (10 of 26) of IDH-mutant glioma patients, whereas no IDH mutation was detected in the normal SVZ. Furthermore, by analyzing cell-type specific mutations, the direction of clonal evolution, the single-cell transcriptome from patient brains and novel mouse model of IDH-mutant glioma arising from mutation-carrying OPCs, we determined that GPCs, including OPCs, harboring the initial driver mutation are responsible for the development and evolution of IDH-mutant gliomas. In summary, our results demonstrate that GPCs containing the IDH mutation are the cells-of-origin harboring the initial driver mutation in IDH-mutant gliomas. ### Competing Interest Statement J.H.L. is a co-founder and chief scientific officer of Sovargen Inc. Y.S.J. is a co-founder and chief executive officer of Inocras Inc. The remaining authors declare no competing interests.

IntroductionThe prognosis within each subtype varies due to histological and molecular factors. This study leverages omics datasets and machine learning to identify biomarkers associated with EC recurrence in different molecular subtypes.MethodsUtilizing DNA methylation, RNA-sequencing, and common variant data from 116 EC samples in The Cancer Genome Atlas (TCGA), differentially expressed genes (DEGs) and differentially methylated regions (DMRs) were identified using t-tests between recurrence and non-recurrence groups. These were visualized through volcano plots and heat maps, while decision trees and random forests classified and stratified the samples.ResultsA machine learning analysis combined with box plots showed that in the copy number-high (CN-H) recurrence group, PARD6G-AS1 had decreased methylation, CSMD1 had increased methylation, and TESC expression was higher than the non-recurrence group. In the copy number-low (CN-L) recurrence group, CD44 expression was elevated. Further validation using TCGA clinical data confirmed PARD6G-AS1 hypomethylation and CD44 overexpression as significant indicators of recurrence (p=0.006 and p=0.02, respectively), and both were linked to advanced stage and lymph node metastasis.ConclusionThe study concludes that PARD6G-AS1 hypomethylation and CD44 overexpression are potential predictors of recurrence in CN-H and CN-L EC patients, respectively.

Accurate prediction of the likelihood of recurrence is important in the selection of postoperative treatment for patients with early-stage breast cancer. In this study, we investigated whether deep learning algorithms can predict patients' risk of recurrence by analyzing the pathology images of their cancer histology.We analyzed 125 hematoxylin and eosin-stained whole slide images (WSIs) from 125 patients across two institutions (National Cancer Center and Korea University Medical Center Guro Hospital) to predict breast cancer recurrence risk using deep learning. Sensitivity reached 0.857, 0.746, and 0.529 for low, intermediate, and high-risk categories, respectively, with specificity of 0.816, 0.803, and 0.972, and a Pearson correlation of 0.61 with histological grade. Class activation maps highlighted features like tubule formation and mitotic rate, suggesting a cost-effective approach to risk stratification, pending broader validation. These findings suggest that deep learning models trained exclusively on hematoxylin and eosin stained whole slide images can approximate genomic assay results, offering a cost-effective and scalable tool for breast cancer recurrence risk assessment. However, further validation using larger and more balanced datasets is needed to confirm the clinical applicability of our approach.

Cutaneous squamous cell carcinoma (SCC) is known for its stepwise progression from healthy skin to premalignant actinic keratosis (AK), followed by a malignant transformation to SCC. Unfortunately, less attention has been paid to changes in gene expression in the tumour microenvironment during this process. We retrospectively selected early-stage cutaneous SCC tissue samples containing both invasive and premalignant portions and conducted a spatial transcriptomic experiment using a NanoString GeoMx Digital Spatial Profiler (DSP). First, we selected invasive and premalignant regions of interest (ROIs) for each tissue. We then compared the gene expression patterns between the two portions (invasive versus premalignant) of the three segments: tumour cells, immune cells and fibroblasts, in each ROI. As a result, early-stage cutaneous SCC tissue samples from 17 patients were selected for this study. We identified 29, 14 and 15 differentially expressed genes (DEGs) between the invasive and premalignant portions of the tumour cells, immune cells and fibroblasts, respectively. The top three genes with the highest absolute log2 fold-change were CCDC88C, GJD3 and COMP in tumour cells; SVEP1, TSLP and PPP2R5C in immune cells; and SPAG6, PPP1CA and CCDC68 in fibroblasts. Notably, several genes, such as COMP, SVEP1 and SPAG6, have been linked to the development and function of cancer-associated fibroblasts. Functional enrichment analysis revealed that several pathways were altered in tumour and immune cells. In conclusion, distinctive changes in gene expression patterns were observed as AK progressed to SCC.

Lymphoma is an uncommon type of breast malignancy, with low prevalence. The ultrasonographic findings of breast lymphoma have been described as nonspecific. Breast lymphoma most commonly appears as a solitary hypoechoic mass on US, and usually shows hypervascularity on color Doppler US. Herein, we report an unusual case of breast lymphoma that presented as multiple bilateral hyperechoic nodules on US.

Background The relationship between human papillomavirus (HPV) and Bowen disease (BD) is not fully understood. Objectives To investigate the differences in HPV detection rates in BD samples across various body regions and analyse the expression patterns of p53, p16 and Ki-67 in relation to HPV presence. Methods Tissue samples from patients diagnosed with BD, confirmed through histopathology, were retrospectively collected. Next-generation sequencing was used for HPV DNA detection. Immunohistochemistry (IHC) for p16, p53 and Ki-67 was performed. Results Out of 109 patients with BD, 21 (19.3%) were HPV-positive. All identified types were alpha-HPVs, with HPV-16 being the most common. The HPV detection rate was significantly higher in the pelvic (9/13, 69%, P < 0.001) and digital (5/10, 50%, P = 0.02) areas compared with those in the other regions. HPV presence was significantly correlated with p53 negativity (P = 0.002), the p53 'non-overexpression' IHC pattern (P < 0.001) and p16-p53 immunostain pattern discordance (P < 0.001). Conversely, there was no notable association between HPV presence and p16 positivity, the p16 IHC pattern or Ki-67 expression. Conclusions Our findings suggest the oncogenic role of sexually transmitted and genito-digitally transmitted alpha-HPVs in the pathogenesis of BD in pelvic and digital regions. [Graphical Abstract]

Accurately segmenting cancer lesions is essential for effective personalized treatment and enhanced patient outcomes. We propose a multi-resolution selective segmentation (MurSS) model to accurately segment breast cancer lesions from hematoxylin and eosin (H&E) stained whole-slide images (WSIs). We used The Cancer Genome Atlas breast invasive carcinoma (BRCA) public dataset for training and validation. We used the Korea University Medical Center, Guro Hospital, BRCA dataset for the final test evaluation. MurSS utilizes both low- and high-resolution patches to leverage multi-resolution features using adaptive instance normalization. This enhances segmentation performance while employing a selective segmentation method to automatically reject ambiguous tissue regions, ensuring stable training. MurSS rejects 5% of WSI regions and achieves a pixel-level accuracy of 96.88% (95% confidence interval (CI): 95.97–97.62%) and mean Intersection over Union of 0.7283 (95% CI: 0.6865–0.7640). In our study, MurSS exhibits superior performance over other deep learning models, showcasing its ability to reject ambiguous areas identified by expert annotations while using multi-resolution inputs.

Background:Segmentectomy is a type of limited resection surgery indicated for patients with very early-stage lung cancer or compromised function because it can improve quality of life with minimal removal of normal tissue. For segmentectomy, an accurate detection of the tumor with simultaneous identification of the lung intersegment plane is critical. However, it is not easy to identify both during surgery. Here, the authors report dual-channel image-guided lung cancer surgery using renally clearable and physiochemically stable targeted fluorophores to visualize the tumor and intersegmental plane distinctly with different colors; cRGD-ZW800 (800 nm channel) targets tumors specifically, and ZW700 (700 nm channel) simultaneously helps discriminate segmental planes. Methods:The near-infrared (NIR) fluorophores with 700 nm and with 800 nm channels were developed and evaluated the feasibility of dual-channel fluorescence imaging of lung tumors and intersegmental lines simultaneously in mouse, rabbit, and canine animal models. Expression levels of integrin alpha v beta 3, which is targeted by cRGD-ZW800-PEG, were retrospectively studied in the lung tissue of 61 patients who underwent lung cancer surgery. Results:cRGD-ZW800-PEG has clinically useful optical properties and outperforms the FDA-approved NIR fluorophore indocyanine green and serum unstable cRGD-ZW800-1 in multiple animal models of lung cancer. Combined with the blood-pooling agent ZW700-1C, cRGD-ZW800-PEG permits dual-channel NIR fluorescence imaging for intraoperative identification of lung segment lines and tumor margins with different colors simultaneously and accurately. Conclusion:This dual-channel image-guided surgery enables complete tumor resection with adequate negative margins that can reduce the recurrence rate and increase the survival rate of lung cancer patients.

PURPOSE:Notable effectiveness of trastuzumab deruxtecan in patients with human epidermal growth factor receptor 2 (HER2)-low advanced breast cancer (BC) has focused pathologists' attention. We studied the incidence and clinicopathologic characteristics of HER2-low BC, and the effects of immunohistochemistry (IHC) associated factors on HER2 IHC results. MATERIALS AND METHODS:The Breast Pathology Study Group of the Korean Society of Pathologists conducted a nationwide study using real-world data on HER2 status generated between January 2022 and December 2022. Information on HER2 IHC protocols at each participating institution was also collected. RESULTS:Total 11,416 patients from 25 institutions included in this study. Of these patients, 40.7% (range, 6.0% to 76.3%) were classified as HER2-zero, 41.7% (range, 10.5% to 69.1%) as HER2-low, and 17.5% (range, 6.7% to 34.0%) as HER2-positive. HER2-low tumors were associated with positive estrogen receptor and progesterone receptor statuses (p < 0.001 and p < 0.001, respectively). Antigen retrieval times (≥ 36 minutes vs. < 36 minutes) and antibody incubation times (≥ 12 minutes vs. < 12 minutes) affected on the frequency of HER2 IHC 1+ BC at institutions using the PATHWAY HER2 (4B5) IHC assay and BenchMark XT or Ultra staining instruments. Furthermore, discordant results between core needle biopsy and subsequent resection specimen HER2 statuses were observed in 24.1% (787/3,259) of the patients. CONCLUSION:The overall incidence of HER2-low BC in South Korea concurs with those reported in previously published studies. Significant inter-institutional differences in HER2 IHC protocols were observed, and it may have impact on HER2-low status. Thus, we recommend standardizing HER2 IHC conditions to ensure precise patient selection for targeted therapy.

Abstract The nanoString GeoMx® Digital Spatial Profiling (DSP) enables the investigation of spatial assessment of tumors through high-plex profiling at the RNA and protein levels. Recent genomic analyses have revealed the intertumor heterogeneity between primary and metastatic lesions in colorectal cancer (CRC) patients. In CRC with liver metastases, current treatment strategies are mainly based on the parameters of primary tumors, and metastatic heterogeneity is a challenge since molecular heterogeneity contributes to therapeutic resistance. Furthermore, spatial intratumor heterogeneity exists within a single tumor between cancer cells (tumor) and their microenvironment (stroma) in human cancers. However, whether there are distinct spatial gene expression patterns of tumor and stroma between primary lesions and liver metastases remains unclear in CRCs. We examined 24 formalin-fixed paraffin-embedded tissue samples, including primary lesions and matched liver metastases from 12 patients (6 with synchronous and 6 with metachronous metastases) using NanoString GeoMx® DSP. The best regions of interest (ROI) with the invasive front boundary of the tumor were selected by pathologists. The ROI was segmented into PanCK-positive (tumor) and PanCK-negative (stroma), followed by a collection of indexed oligonucleotides and sequencing on Illumina instrument. Differential expression and pathways enrichment analyses were performed using R BioConductor package standR, limma, and GSEABase. Statistical significances were based on |Log2 fold change| > 1 and Benjamini-Hochberg-corrected P < 0.05. Immune cell abundance was estimated using the SpatialDecon package. In 12 metastatic patients (mean age, 61 years ± 7 [standard deviation]), liver metastatic stroma was associated with 68 upregulated and 93 downregulated genes compared to the primary stroma, with enrichment of 'Immune response' terms in gene set enrichment analysis (GSEA). A higher proportion of CD4 T memory cells and increased expression of cytotoxic genes (IL7R [2.02-fold], CD3E [1.75-fold], KLRB1 [1.68-fold], and GZMK [1.54-fold]) were observed in liver stroma compared to primary stroma. Conversely, liver metastatic tumors were associated only with 23 upregulated and 6 downregulated genes compared to primary tumors, with 'humoral response' terms enriched in GSEA. Similar patterns were observed in stratified analyses with synchronous and metachronous metastases, although a more prominent ‘immune response’ was observed in synchronous metastatic stroma compared to the primary stroma. We revealed that higher immune cell proportion and cytotoxic activity were observed in liver metastatic stroma compared to CRC stroma, providing novel insights into a unique etiology and may yield clinical implications for developing targeted treatment modalities for liver metastatic CRC patients. Citation Format: Jongwon Lee, Yeseul Kim, Hyo Seon Ryu, Jongmin Sim, Chungyeul Kim, Jong Min Park, Ah-Reum Lim, Jung Sun Kim, Hwa Jung Sung, Xingyi Guo, Jungmin Choi, Jungyoon Choi. Investigating spatial gene expression profiling in colorectal cancer: Tumor and stroma comparison between primary lesions and matched liver metastases [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 1160.

Background: Amplification of the 3q region has been identified as a useful biomarker for the diagnosis and treatment of squamous cell carcinoma (SqCC). This region contains genes such as PIK3CA and YEATS2, which have been linked to the prognosis of SqCC.Methods: The NanoString nCounter assay is a powerful tool for identifying genetic alterations that affect the progression and prognosis of SqCC. The NanoString nCounter assay was used to identify a subgroup of patients with gene level gain in the 3q region.Results: Gene level gain in the 3q region was more frequent in SqCC than in adenocarcinoma. We found that genes such as PIK3CA and YEATS2 in the 3q region were associated with the prognosis of SqCC. Therefore, identifying a subgroup of patients with gene level gain in the 3q region using the NanoString nCounter assay can aid in selecting appropriate treatment options and improving prognostic predictions for SqCC patients.Conclusion: Amplification of the 3q region in SqCC of lung cancer is a useful biomarker for diagnosis and treatment. The NanoString nCounter assay is a powerful tool for identifying specific genetic alterations that affect the progression and prognosis of SqCC. Our study highlights the importance 3q amplification and its associated genes in lung cancer.

CCR Translation for This Article from Cytoplasmic Estrogen Receptor in Breast Cancer

Abstract Purpose: In addition to genomic signaling, it is accepted that estrogen receptor-α (ERα) has nonnuclear signaling functions, which correlate with tamoxifen resistance in preclinical models. However, evidence for cytoplasmic ER localization in human breast tumors is less established. We sought to determine the presence and implications of nonnuclear ER in clinical specimens. Experimental Design: A panel of ERα-specific antibodies (SP1, MC20, F10, 60c, and 1D5) was validated by Western blot and quantitative immunofluorescent (QIF) analysis of cell lines and patient controls. Then eight retrospective cohorts collected on tissue microarrays were assessed for cytoplasmic ER. Four cohorts were from Yale (YTMA 49, 107, 130, and 128) and four others (NCI YTMA 99, South Swedish Breast Cancer Group SBII, NSABP B14, and a Vietnamese Cohort) from other sites around the world. Results: Four of the antibodies specifically recognized ER by Western and QIF analysis, showed linear increases in amounts of ER in cell line series with progressively increasing ER, and the antibodies were reproducible on YTMA 49 with Pearson correlations (r2 values) ranging from 0.87 to 0.94. One antibody with striking cytoplasmic staining (MC20) failed validation. We found evidence for specific cytoplasmic staining with the other four antibodies across eight cohorts. The average incidence was 1.5%, ranging from 0 to 3.2%. Conclusions: Our data show ERα is present in the cytoplasm in a number of cases using multiple antibodies while reinforcing the importance of antibody validation. In nearly 3,200 cases, cytoplasmic ER is present at very low incidence, suggesting its measurement is unlikely to be of routine clinical value. Clin Cancer Res; 18(1); 118–26. ©2011 AACR.

Correct prediction of cancer molecular subtype plays a very important role in determining the treatment of cancer patients. In determining the molecular subtype of breast cancer, protein receptors such as estrogen receptors (ER), progesterone receptors (PR), and human epidermal growth factor receptor type 2 (HER2) are used as key factors. These protein receptors also play an important role in determining the treatment method or predicting the prognosis of breast cancer. To confirm this, a test should be performed by immunohistochemical (IHC) staining. In this study, we investigated the morphological relationship between the molecular subtypes and hematoxylin and eosin (H&E) stained whole slide images (WSIs) without IHC stained images, and performed an experiment to evaluate that the protein receptors can be predicted by these morphological features. The TCGA-BRCA data were utilized in this study. There were 728 cases out of the total 1097 cases where the IHC status for ER, PR, and HER2 was either positive or negative. Each individual case is scanned using different scanners and magnifications. In some cases with multiple slides, we use only the first scanned image. The entire WSI was randomly split into 3:1:1, and used for training, tuning, and test, respectively. Individual WSIs was tiled into 1024 × 1024 patches for this study. The multi-task deep learning model predicts the protein receptor status of individual patches. In this study, a confidence measure was added to remove the uncertainty of the deep learning model. When learning WSIs, patches of less than 70% based on these confidence scores did not affect WSIs. Specifically, when learning morphological features, we used strong augmentation such as grayscale, gaussian blur, color jitter, and posterization to ensure that predictions were not biased by color alone. Based on individual patches, ER accuracy of 76%, PR accuracy of 65%, and HER2 accuracy of 79% are shown. As a result of prediction by majority voting on WSI images through patch predicted results, ER accuracy of 74.6% PR accuracy of 66%, and HER2 accuracy of 76.6% Through this study, it was confirmed that the sufficiently trained deep learning model predicted ER, PR, and HER2, which are important factors of molecular subtype, relatively well. Through this, it was found that there was some correlation between morphological characteristics and molecular subtypes in H&E stained WSIs. If more data are collected through future experiments, a molecular subtype prediction model can be developed. Citation Format: Geongyu Lee, Chungyeul Kim, Tae-Yeong Kwak, Sun Woo Kim, Hyeyoon Chang. Predicting protein receptor status from H&E-stained images in breast 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 5404.

PDF file - 81K, ColoCarta panel