Large-scale genomic analysis of pancreatic cancer in a real-world patient population

Pancreatic cancer (PC) remains one of the deadliest malignancies, with minimal improvement in survival over recent decades. Fundamental questions persist regarding how genomic alterations influence PC biology and therapeutic response. Using a large multi-modal real-world genomic database, we sought to characterize the mutational and transcriptional landscape of PC and identify patterns that may inform future precision medicine approaches. Whole-exome sequencing (WES) data from tumor tissue samples were analyzed for patients with PC included in Natera's proprietary Real-World Database. Sequencing was performed as part of the tumor-informed, personalized Signatera™ circulating tumor DNA assay designed between May 2019-Nov 2024. After excluding cases of low tumor mutational burden and variant allele frequency, a total of 3,664 patients were included in the analysis. Variant calling was performed using Mutect2 and VarScan2. For prevalence analysis, only non-synonymous somatic SNVs and INDELs were included. Germline variant analysis was performed for 128 cancer predisposition genes in both tumor and normal samples; ClinVar pathogenic/likely pathogenic (P/LP) germline variants were retained. RNA sequencing data, from Altera TM tumor genomic profiling test (Natera, Inc.), were available for 658 patients, of whom 391 had matched WES data. Among the 3,664 patients with PC, the male/female ratio was [50%/50%], and the stage distribution was: I (14.6%%), II (24.4%), III (24.8%), IV (28.2%), and unknown (7.9%). Genetic ancestry composition included European (74.5%), African (10.9%), Latino/Admixed American (7.9%), East Asian (5.0%), and South Asian (1.5%) populations. The most frequently somatically mutated genes were KRAS (78%), TP53 (62%), CDKN2A (18%), SMAD4 (18%), and ARID1A (8%). The most common individual somatic variants were KRAS G12D (32%), KRAS G12V (25%), KRAS G12R (13%), TP53 R175H (5%), and KRAS Q61H (4%). No statistically significant ancestry-based differences were observed in either gene or variant-level frequencies. P/LP germline variants were detected in 210 patients, comprising 149 unique variants across 34 genes, most commonly ATM , BRCA2 , MUTYH , BRCA1 , and PALB2 . Among RNA-Seq profiled tumors, expression analysis identified patients with basal and classical molecular subtypes associated with distinct genomic features and clinical outcomes. This real-world study represents one of the most extensive characterizations of PC to date, integrating WES, germline, and RNA-Seq data from over 3,500 patients. The findings confirm the predominance of canonical driver alterations ( KRAS , TP53 , CDKN2A , SMAD4 ) and highlight germline and transcriptomic diversity across the disease spectrum. Ongoing analyses are assessing transcriptional variation across KRAS and other genomic features and will link these profiles with survival outcomes to determine clinical relevance.