To identify gut and oral metagenomic signatures that accurately predict pancreatic ductal carcinoma (PDAC) and to validate these signatures in independent cohorts. We conducted a multinational study and performed shotgun metagenomic analysis of fecal and salivary samples collected from patients with treatment-naïve PDAC and non-PDAC controls in Japan, Spain, and Germany. Taxonomic and functional profiles of the microbiomes were characterized, and metagenomic classifiers to predict PDAC were constructed and validated in external datasets. Comparative metagenomics revealed dysbiosis of both the gut and oral microbiomes and identified 30 gut and 18 oral species significantly associated with PDAC in the Japanese cohort. These microbial signatures achieved high area under the curve values of 0.78 to 0.82. The prediction model trained on the Japanese gut microbiome also had high predictive ability in Spanish and German cohorts, with respective area under the curve values of 0.74 and 0.83, validating its high confidence and versatility for PDAC prediction. Significant enrichments of Streptococcus and Veillonella spp and a depletion of Faecalibacterium prausnitzii were common gut signatures for PDAC in all the 3 cohorts. Prospective follow-up data revealed that patients with certain gut and oral microbial species were at higher risk of PDAC-related mortality. Finally, 58 bacteriophages that could infect microbial species consistently enriched in patients with PDAC across the 3 countries were identified. Metagenomics targeting the gut and oral microbiomes can provide a powerful source of biomarkers for identifying individuals with PDAC and their prognoses. The identification of shared gut microbial signatures for PDAC in Asian and European cohorts indicates the presence of robust and global gut microbial biomarkers. Display omitted This study analyzed the gut and oral microbiomes of patients with pancreatic cancer and showed the potential of the microbial species to be used as a biomarker for the disease and its prognosis.