BACKGROUND: high recurrence rates of up to 75% within 2 years in pancreatic ductal adenocarcinoma (PDAC) patients resected for cure indicate a high medical need for clinical prediction tools and patient specific treatment approaches. Addition of the EGFR inhibitor erlotinib to adjuvant chemotherapy failed to improve outcome but its efficacy in some patients warrants predictors of responsiveness. PATIENTS AND METHODS: we analysed tumour samples from 293 R0-resected patients from the randomized, multicentre phase III CONKO-005 trial (gemcitabine ± erlotinib) with targeted sequencing, copy number, and RNA expression analyses. FINDINGS: a total of 1086 mutations and 4157 copy-number aberrations (CNAs) with a mean of 17.9 /tumour were identified. Main pathways affected by genetic aberrations were the MAPK-pathway (99%), cell cycle control (92%), TGFβ signalling (77%), chromatin remodelling (71%), and the PI3K/AKT pathway (65%). Based on genetic signatures extracted with non-negative matrix factorization we could define five patient clusters, which differed in mutation patterns, gene expression profiles, and survival. In multivariable Cox regression analysis, SMAD4 aberrations were identified as a negative prognostic marker in the gemcitabine arm, an effect that was counteracted when treated with erlotinib (DFS: HR=1.59, p = 0.016, and OS: HR = 1.67, p = 0.014). Integration of differential gene expression analysis established SMAD4 alterations with low MAPK9 expression (n = 91) as a predictive biomarker for longer DFS (HR=0.49; test for interaction, p = 0.02) and OS (HR = 0.32; test for interaction, p = 0.001). INTERPRETATION: this study identified five biologically distinct patient clusters with different actionable lesions and unravelled a previously unappreciated association of SMAD4 alteration status with erlotinib effectiveness. Confirmatory studies and mechanistic experiments are warranted to challenge the hypothesis that SMAD4 status might guide addition of erlotinib treatment in early-stage PDAC patients.