Hypoxia is common in non-small cell lung cancer (NSCLC) and an attractive therapeutic target. As hypoxia-targeting treatments are effective in patients with the most hypoxic tumours, we aimed to develop a lung adenocarcinoma (LUAD) hypoxia-related gene expression signature. RNAseq was used to identify genes significantly differentially expressed under hypoxia (1% O2) in four LUAD cell lines. Identified genes were used for unsupervised clustering of a TCGA-LUAD training dataset (n = 252) and in a machine learning approach to build a hypoxia-related signature. Thirty-five genes were upregulated in common in three of the four lines and reduced in the training cohort to a 28-gene signature. The signature was prognostic in the TCGA training (HR 2.12, 95% CI 1.34-3.37, p = 0.0011) and test (n = 250; HR 2.13, 95% CI 1.32-3.45, p = 0.0016) datasets. The signature was prognostic for overall survival in a meta-analysis of nine other datasets (n = 1257; HR 2.08, 95% CI 1.60-2.70, p < 0.0001). The 28-gene LUAD hypoxia related signature can be taken forward for further validation using a suitable gene expression platform.
© 2022. The Author(s).