Abstract 753: AI-powered deep visual proteomics (DVP) for early pancreatic cancer insights

Abstract Pancreatic ductal adenocarcinoma (PDAC), the most common form of pancreatic cancer, presents a formidable clinical challenge. Patients experience 70-80% post-resection mortality due to recurrence. Identifying aberrant molecular pathways and drivers of pre-cancer progression is an unmet area of importance for developing effective biomarkers and enabling early detection. Pancreatic intraepithelial neoplasia (PanINs) represent the key precursor to PDAC. While genomic and transcriptomic profiling of these lesions has been performed at depth, the deep proteomic profiles of these precursors remains largely unexplored. Deep Visual Proteomics (DVP) combines advanced microscopy, machine learning-based image analysis, automated laser microdissection, and high-sensitivity mass spectrometry to analyze specific cell populations while maintaining their spatial context within tissue samples. Here, we combine DVP with cutting-edge pathology foundation models trained on over 600,000 histopathological slides and data from more than 200,000+ patients, to systematically and precisely characterize key regions of interest (ROIs) in incidental panIN lesions in organ donor samples. We then employ our spatially-resolved mass-spectrometry based proteomics pipeline to compare these pancreatic precursor lesions, including acinar ductal metaplasia and regions of low- and high-grade PanINs, to normal healthy ducts. Hereby we developed the first comprehensive spatial proteomics maps of pancreatic precursor lesions. This AI-powered ROI classification, coupled with our advanced DVP platform, delivers unprecedented molecular insights. Our platform achieves remarkable proteomic depth, identifying ∼6,000 proteins from ∼100 phenotype-matched cells per sample, revealing stage-specific molecular signatures in PanIN lesions. This first-of-its-kind dataset enables the creation of detailed, phenotype-specific and spatially-resolved proteomic landscapes of PDAC progression. Our innovative approach reveals progressive activation of KRAS signaling across the PanIN-to-PDAC transition, corroborating key drivers of multistep carcinogenesis. Furthermore, we identify several promising biomarkers candidates, including targets that might form the basis for early detection and also provide unprecedented insights into targets for cancer interception. Combining the precision of AI-based pathology with the depth of DVP represents a paradigm shift in the toolkit to study PDAC progression, and generates novel molecular insights in panIN lesions that may enable meaningful intervention in pancreatic cancer. Citation Format: Jimin Min, Lisa Schweizer, Gijs Zonderland, Benson Chellakkan Selvanesan, Lukas Oldenburg, Michael A. Hollingsworth, Paul M. Grandgenett, Ishani Ummat, Maximilian T. Strauss, Andreas Mund, Anirban Maitra. AI-powered deep visual proteomics (DVP) for early pancreatic cancer insights [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 753.