Abstract B124: Rapid, high-plex spatial proteomic profiling of pancreatic cancer using a pancreas-focused-assay and customized analysis workflow

Abstract INTRODUCTION: Pancreatic ductal adenocarcinoma (PDAC) is currently the 3rd leading cause of cancer-related deaths in the United States. While studying tissue from human patients holds promise for developing new therapies, this approach is challenging due to the low neoplastic cell content of pancreatic cancer and the availability of small biopsy specimens as the only tissue source for patients with advanced disease. To maximize the amount of information that can be obtained from such specimens, we constructed a high-plex spatial proteomics assay specifically designed for PDAC and developed a matched computational analysis workflow that delivers rapid results suitable for correlative science studies. METHODS: PULSAR (Pancreatic cancer Unified Landscape Spatial Architecture Report) is a comprehensive 61-plex protein assay designed for the PhenoCycler platform (Akoya Biosciences) that provides highly granular, single-cell, spatially-resolved characterization of PDAC, including tumor basal/classical subtype, driver gene status, immune cell infiltration and immune checkpoint expression, stromal fibroblast composition, and cell proliferation status. Antibodies (including 21 pre-conjugated and 40 custom-conjugated) were selected for their known relevance to PDAC biology and were optimized using primary PDAC specimens. To process PULSAR digital image data, we used machine learning-based algorithms as implemented in the QuPath software package to perform tissue segmentation, cell segmentation, and initial cell phenotyping. Additional downstream processing was performed using CELLestial, a novel end-to-end pipeline that performs cell phenotyping using automated thresholding, downstream analyses and data visualization in a custom-scripted interface. RESULTS: Across 30 primary pancreatic cancer specimens, the average time for tissue preparation, processing, and image collection was 63 hours, with four cases processed during a week and 100% of images yielding analyzable data. Overall, the per marker staining success rate was 90%, with 90% of all cases having >90% of all markers successfully imaged. Average image size was 58.7 GB, and the number of phenotyped cells ranged from 510,474—957,964 per case. Image processing using QuPath and CELLestial required 2 days on average, including initial quality control review, algorithm training, and refinement of cell phenotypes. Across the 61 markers, more than 75 cell phenotypes were automatically identified, and more than 500 specimen-level features, such as density for a given immune cell type and tumor basal/classical subtype abundance, are calculated. CONCLUSION: By combining a purpose-built assay with a customized bioinformatic workflow, we were able to conduct 61-plex protein-based profiling of pancreatic cancer specimens with a high per-marker success rate in short time frame. Our approach provides a comprehensive assessment of pancreatic cancer and its microenvironment and can accelerate biomarker discovery and therapeutic innovation for PDAC. Citation Format: Andressa Dias Costa, Emma C. Coleman, Audrey P. Lacy, Carla Rujana, Dalia Elganainy, Micaela Morgado, Cibelle F. Lima, Canh Hiep Nguyen, C.Sloane Furniss, Andrew J. Aguirre, Brian M. Wolpin, Simona Cristea, Jonathan A. Nowak. Rapid, high-plex spatial proteomic profiling of pancreatic cancer using a pancreas-focused-assay and customized analysis workflow [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Advances in Pancreatic Cancer Research—Emerging Science Driving Transformative Solutions; Boston, MA; 2025 Sep 28-Oct 1; Boston, MA. Philadelphia (PA): AACR; Cancer Res 2025;85(18_Suppl_3):Abstract nr B124.