Robust identification of an aneuploidy-specific tumor signature in PDAC through integrated single-cell ploidy inference and multi-algorithmic consensus screening

Chromosomal instability (CIN)-driven aneuploidy promotes intratumoral heterogeneity and treatment resistance in pancreatic ductal adenocarcinoma (PDAC), but the transcriptional programs accompanying the diploid-to-aneuploid transition and their effects on the tumor microenvironment remain incompletely defined. We analyzed 52,072 single cells from six paired PDAC and adjacent non-tumor specimens (GSE212966). CopyKAT-based ploidy inference and high-dimensional weighted gene co-expression network analysis were combined with a six-algorithm consensus machine-learning pipeline using a 70:30 training-test split. Prognostic performance was evaluated in TCGA-PAAD and three independent GEO cohorts. CellChat was used to characterize tumor-microenvironment communication, and the functional role of PLAT was tested by siRNA/shRNA knockdown, EdU and Transwell assays, and a mouse xenograft model. A four-gene High-confidence Tumor Signature (HTS; CEACAM6, SFN, GPRC5A, and PLAT) was identified and increased progressively along the inferred diploid-to-aneuploid trajectory. In the independent test set, the combined HTS distinguished aneuploid from diploid ductal cells with an AUC of 0.939. HTS-high malignant cells displayed enhanced LAMININ, COLLAGEN, and MIF signaling. In TCGA-PAAD, the HTS was independently associated with overall survival (HR = 1.83, 95% CI 1.19–2.80, P = 0.006) and improved the clinical-model C-index by 0.0374; risk stratification was reproduced in all three external GEO cohorts. PLAT knockdown reduced PDAC-cell proliferation and invasion in vitro and suppressed xenograft growth in vivo. The HTS provides a robust transcriptional readout of aneuploid malignant states in PDAC, links chromosomal instability to specific microenvironmental signaling, and identifies PLAT as a functionally validated candidate therapeutic target.