Multi-omics integration and machine learning uncover molecular basal-like subtype of pancreatic cancer and implicate A2ML1 in promoting tumor epithelial-mesenchymal transition

作者：Jiachen Ge, Jian‐Ping Cai, Gaolei Zhang, Deyu Li, Lianyuan Tao · 发表于：Journal of Translational Medicine · 年份：2025 · DOI：10.1186/s12967-025-06711-z · 被引用次数：13 · 研究领域：Pancreatic and Hepatic Oncology Research、NF-κB Signaling Pathways、Chromatin Remodeling and Cancer

BACKGROUND: Pancreatic cancer (PC) represents a highly heterogeneous malignancy with poor prognosis, where precise molecular subtyping facilitates comprehensive understanding of disease progression. METHODS: Transcriptomic, methylation, and mutational data from 168 PC samples were integrated. The molecular subtypes were established and validated across 13 independent cohorts utilizing ten distinct classification methods. Prognostic genes were identified to construct predictive models through various machine learning approaches. Following the identification of A2ML1 as a key gene, its expression profile was detected using RT-qPCR, western blotting, and immunohistochemistry. We subsequently conducted functional experiments to elucidate the mechanism of A2ML1 in promoting PC progression. RESULTS: Through multi-omics integration, we classified PC into two molecular subtypes with distinct prognostic outcomes. Pathway enrichment analysis revealed subtype-specific characteristics, subsequently validated in external cohorts. Using 23 prognostic genes, we developed and validated a prognostic signature through 101 machine learning algorithms and their combinations, with ridge regression demonstrating optimal performance. This signature demonstrated superior accuracy compared to multiple published signatures. The risk scores showed significant correlations with drug sensitivity, clinical characteristics, and patient outcomes. We further validated that A2ML1 expression was significantly ...