AI-Powered Deep Visual Proteomics Reveals Critical Molecular Transitions in Pancreatic Cancer Precursors
Date: | July 1, 2026 |
PMID: | |
Category: | N/A |
Authors: | Jimin Min, Lisa Schweizer, Gijs Zonderland, Benson Chellakkan Selvanesan, Julie H Thomsen, Lukas Oldenburg, Seong-Woo Bae, Bongjun Kim, Sharía D Hernández, Gabriela Jez, Vincent Bernard, Benjamin J Swanson, Kelsey A Klute, Huamin Wang, Thomas C Caffrey, Paul M Grandgenett, Michael A Hollingsworth, Ishani Ummat, Maximilian T Strauss, Andreas Mund, Anirban Maitra |
Abstract: |
|
Pancreatic ductal adenocarcinoma (PDAC) evolves through precursors, yet the protein programs governing early progression remain poorly defined. We applied Deep Visual Proteomics (DVP)-integrating computational pathology, laser microdissection, and mass spectrometry (MS)-to profile normal ducts, acinar-to-ductal metaplasia (ADM), low-grade (LG) and high-grade (HG) pancreatic intraepithelial neoplasia (PanIN), and invasive carcinoma from organ donors and patients with PDAC. Quantifying 9,181 proteins from ∼100 cells per region, we uncovered a molecular field effect in histologically normal ducts and proteomic divergence of LG-PanINs by cancer context. We identified four stage-associated molecular programs. Stress adaptation and immune engagement emerged early in cancer-associated normal ducts. Metabolic reprogramming initiated in normal ducts and intensified across PanIN progression. Mitochondrial remodeling became prominent in HG-PanINs before invasion. MS detected KRAS hotspot mutant peptides within incidental precursor lesions from cancer-free individuals. These findings demonstrate that molecular reprogramming precedes histologic transformation, creating opportunities for earlier detection of lethal cancer.
Significance: Artificial intelligence (AI)-guided DVP represents the first in-depth assessment of the proteomic landscapes observed during the multistep progression of pancreatic adenocarcinoma, including histologically normal ducts, ADM, and LG- and HG-PanIN lesions. These data represent a unique resource of candidate biomarkers and interception targets against this lethal disease. See related commentary by Yang and Fan, p. 1255.
Acknowledgements:
The content of this article is solely the responsibility of the authors and does not necessarily represent the official views of the National Cancer Institute, or the National Institute of Health.
The Translational and Basic Science Research in Early Lesions (TBEL) Research Consortia is supported and funded by grants from the National Cancer Institute and the National Institutes of Health under the following award numbers:
Project Number: | Awardee Organization |
U54CA274374 | Fred Hutchinson Cancer Center |
U54CA274375 | Houston Methodist Research Institute |
U54CA274370 | Johns Hopkins University |
U54CA274371 | UT MD Anderson Cancer Center |
U54CA274367 | Vanderbilt University Medical Center |
Comments