Abstract Background: Pancreatic ductal adenocarcinoma (PDAC) tumor inter-patient heterogeneity has been well described with two major subtypes (classical and basal-like). An important intra-patient heterogeneity has not yet been extensively studied due to the lack of standardized and easily accessible high throughput methods compatible with clinical routine. Material and Methods: To build an immunohistochemical (IHC) tool capable of differentiating classical and basal-like tumors, relevant antibodies were selected using a multi-step process. The successive stages of an in-silico selection from the bulk transcriptome of 309 PDACs and 30 patient-derived xenografts followed by the validation on tissue microarrays in a cohort of PDAC (n=50) were conducted. We used our final IHC panel on whole slides of serial histology sections in 95 resected PDACs. After image registration, we performed a tile-based-analysis in tumoral and pre-neoplastic areas and a k-means clustering to identify relevant marker combinations. Results: Our final selection of the IHC panel included GATA6, CLN18, TFF1, MUC16, S100A2, KRT17, PanBasal. Among the 44,042 tumor tiles analyzed, 4 different phenotypes were identified: 1 classical (CLN18+, TFF1+, GATA6+), 1 intermediate (KRT17+) and 2 basal-like (MUC16+ vs S100A2+) with specific biological properties. A minor basal contingent drastically reduced overall survival (OS) in the multivariate analysis (HR=2.36, p=0.01). In contrast, the presence of the intermediate and classical clusters did not impact OS. However, among patients with PDAC expressing the classical cluster, a tremendous difference of OS was observed according to the co-occurence of a basal cluster (HR=0.34, p=0.003). These prognostic associations have been validated in an independent cohort of 150 patients with PDAC on tissue microarrays. By analyzing preneoplastic lesions, we showed the evolution process of pancreatic carcinogenesis going from a classical toward a basal through an early intermediate phenotype. Conclusion: Our restricted panel of antibodies could help pathologists to predict the high degree of PDAC heterogeneity, ultimately leading to an improvement of patient management. Citation Format: Marc Hilmi, Flore Delecourt, Jerôme Rafenne, Taib Bourega, Nelson Dusetti, Juan Iovanna, Yuna Blum, Magali Richard, Cindy Neuzillet, Anne Couvelard, Louis De Mestier, Vinciane Rebours, Rémy Nicolle, Jérôme Cros. Redefining phenotypic heterogeneity of pancreatic ductal adenocarcinoma: A bottom-up approach [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Pancreatic Cancer; 2023 Sep 27-30; Boston, Massachusetts. Philadelphia (PA): AACR; Cancer Res 2024;84(2 Suppl):Abstract nr B064.