{"id":989,"date":"2025-04-22T00:00:00","date_gmt":"2025-04-22T00:00:00","guid":{"rendered":"https:\/\/www.lunit.io\/media-hub\/lunit-showcases-ai-pathology-advances-in-cancer-research-from-rare-to-common-tumors-at-aacr-2025\/"},"modified":"2025-10-31T15:39:32","modified_gmt":"2025-10-31T15:39:32","slug":"lunit-showcases-ai-pathology-advances-in-cancer-research-from-rare-to-common-tumors-at-aacr-2025","status":"publish","type":"media-hub","link":"https:\/\/www.lunit.io\/en\/media-hub\/lunit-showcases-ai-pathology-advances-in-cancer-research-from-rare-to-common-tumors-at-aacr-2025\/","title":{"rendered":"Lunit Showcases AI Pathology Advances in Cancer Research from Rare to Common Tumors at AACR 2025"},"content":{"rendered":"

\"\"<\/p>\n

SEOUL, South Korea, April 22, 2025<\/strong> \u2014 Lunit (KRX:328130.KQ), a leading provider of AI-powered solutions for cancer diagnostics and therapeutics, today announced the presentation of seven posters at the American Association for Cancer Research (AACR) Annual Meeting 2025, taking place April 25\u201330 in Chicago, Illinois. The presentations introduce Lunit\u2019s latest research in AI-based histopathology, featuring studies powered by the Lunit SCOPE\u00ae suite across a range of cancers\u2014from rare salivary gland tumors to common types like lung cancer.<\/p>\n

The poster lineup includes two collaborative studies with global pharma leaders, including AstraZeneca. One study\u2014conducted in partnership with AstraZeneca\u2014presents the development and validation of an AI model that predicts EGFR mutations from H&E slides, enabling faster, more accessible mutation testing for NSCLC patients. Another, co-authored with a major global biotech company, applies Lunit SCOPE IO\u00ae to phase II and III clinical trial data to predict benefit from atezolizumab, revealing that AI-based histologic profiling can help stratify patients based on likely immunotherapy response.<\/p>\n

Among the seven studies to be presented, three feature high-impact findings with particular clinical and scientific relevance.<\/p>\n

One study addresses the challenge of predicting response to neoadjuvant immuno-chemotherapy in patients with resectable salivary gland cancer (SGC)\u2014a rare and aggressive cancer. To better understand treatment outcomes, the study applied a multi-modal approach that combined single-cell RNA sequencing, T cell receptor (TCR) analysis, spatial transcriptomics (Xenium), and AI-powered histological profiling using Lunit SCOPE IO\u00ae on surgically resected tumor samples. Responders were found to have more CD8+ dysfunctional and memory T cells, along with increased TCR clonality and reduced diversity\u2014patterns suggestive of clonal expansion. Non-responders, in contrast, showed a higher presence of tumor-associated macrophages. Lunit SCOPE IO\u00ae contributed to detailed morphological profiling, enabling cell type identification and validation of spatial patterns. These findings suggest that combining advanced molecular tools with AI-powered histology may help uncover tumor microenvironment features predictive of response to immunotherapy.<\/p>\n

Another study explores potential biomarkers linked to treatment resistance in salivary duct carcinoma (SDC) by combining Lunit SCOPE IO\u00ae with Xenium spatial transcriptomics. The researchers analyzed over 915,000 cells from surgically resected salivary gland tumors, including SDC cases treated with neoadjuvant immunotherapy. In one relapsed case, the tumor showed higher expression of genes associated with immune evasion and epithelial-to-mesenchymal transition (EMT), despite similar morphology to a non-relapsed case. A lower presence of CXCL9-expressing tumor-infiltrating lymphocytes was also noted, which may reflect a less immunologically active tumor microenvironment. These findings offer additional insight into potential resistance-related features that may not be evident through conventional histology.<\/p>\n

In a third study, Lunit developed an AI model to identify EGFR-mutant NSCLC tumors with morphologic features similar to small cell lung cancer (SCLC)\u2014a pattern clinically linked to early histologic transformation from NSCLC to SCLC and resistance to EGFR tyrosine kinase inhibitors (TKIs), particularly in patients with RB1 mutations. The study used deep learning to analyze H&E-stained tumor slides from 106 advanced-stage EGFR-mutant NSCLC patients, performing cell-level tumor heterogeneity analysis based on AI-discovered morphological features. Patients in the top 25% for SCLC-like morphology\u2014defined as the SCLC-like group\u2014had significantly smaller nuclear area (56 \u00b5m\u00b2 vs. 102 \u00b5m\u00b2) and darker nuclear staining. Clinically, they experienced shorter progression-free survival after TKI therapy and were more likely to later transform into SCLC upon rebiopsy (15.8% vs. 2.0%). This study is the first to demonstrate that AI-based morphologic profiling at diagnosis can identify patients at risk for small cell transformation and early TKI resistance, offering a new path toward risk-adapted treatment planning.<\/p>\n

The remaining studies further demonstrate the breadth of Lunit\u2019s research capabilities and AI expertise. These include studies on cell surface target discovery in prostate cancer and preclinical immunotherapy enhancement in colon cancer.<\/p>\n

\"At AACR 2025, we\u2019re showcasing how Lunit\u2019s AI technologies are driving a new wave of biomarker discovery and clinical insight,\" said Brandon Suh, CEO of Lunit. \"From salivary gland cancer to lung cancer, our studies reveal how AI-powered histopathology\u2014especially through Lunit SCOPE IO\u00ae\u2014can uncover critical tumor microenvironment patterns and transformation risks, and even predict how tumors may respond to targeted therapies well before clinical progression is observed. These insights can play a meaningful role in shaping more precise, responsive cancer care.\"<\/p>\n

To learn more about Lunit\u2019s latest research and activities at AACR 2025, visit Booth #2843<\/strong>.<\/p>\n

Lunit\u2019s featured presentations at AACR 2025 include:<\/strong><\/p>\n

 <\/p>\n