Development and multicohort evaluation of a KRT5/KRT6A/KRT17-based prognostic scoring system for pancreatic cancer using single-cell RNA sequencing

Due to its insidious onset and lack of effective risk stratification tools, pancreatic cancer (PACA) outcomes remain poor, underscoring the urgent need for an optimized prognostic scoring system. The CRA001160 single-cell RNA-sequencing (scRNA-seq) dataset (24 PACA samples) was used for single-cell analyses. A total of 913 bulk samples from four independent cohorts were included for prognostic score development and multicohort evaluation. Multiplex immunofluorescence using a tissue microarray cohort (TMA, n = 80) was utilized to validate the performance of our prognostic scoring system. Analysis of malignant cells in CRA001160 identified a C8 malignant-cell state with proliferation- and stemness-related programs whose abundance increased across the available TNM stages. Across the bulk cohorts, C8-related features were correlated with poorer survival and classic tumor suppressor TP53 and CDKN2A mutation status. Using three markers of the C8 malignant-cell state, we developed and evaluated a novel prognostic scoring system demonstrating distinct survival outcomes. The results of the TMA and four independent transcriptomic cohorts showed consistent prognostic associations, supporting further evaluation across independent datasets and assay types. Our findings provide insights into PACA intratumoral heterogeneity and illustrate the potential of single-cell analyses for prognostic biomarker development. The KRT5 / KRT6A / KRT17 -based score showed prognostic relevance across retrospective cohorts but requires prospective validation using standardized assays and prespecified thresholds before clinical application.