Exploring pancreatic ductal adenocarcinoma (PDAC) metabolic landscape would contribute to further understand PDAC from the metabolic perspective and provide more details for precise treatment design. This study aims to describe metabolic landscape of PDAC. Bioinformatics analysis was used to investigate the differences of genome, transcriptome, and proteome levels of metabolic patterns. Three subtypes (MC1, MC2, and MC3) were identified and characterized as distinct metabolic patterns. MC1, enriched in lipid metabolism and amino acid metabolism signatures, was associated with lower abundance of immune cells and stromal cells, and non-response to immunotherapy. MC2 displayed immune-activated characteristics, minor genome alterations and good response to immunotherapy. MC3 was characterized by high glucose metabolism, high pathological grade, immune-suppressed features, poor prognosis, and epithelial-mesenchymal transition phenotype. A ninety-three gene classifier preformed robust prediction and high accuracy (training set: 93.7%; validation set 1: 85.0%; validation set 2: 83.9%). Using random forest classifier, probabilities of three patterns could be predicted on pancreatic cancer cell lines, which could be used to find vulnerable targets in response to both genetic and drug perturbation. Our study revealed features of PDAC metabolic landscape, which could be expected to provide a reference for prognosis prediction and precise treatment design.
© 2023. The Author(s), under exclusive licence to Springer Nature America, Inc.