Machine learning revealed ferroptosis features and ferroptosis-related gene-based immune microenvironment in lung adenocarcinoma.

Ferroptosis has been identified as a novel type of programmed cell death that has a major effect on the development of lung adenocarcinoma. Nevertheless, there has yet to be a clear set of therapeutic targets based on ferroptosis. This study seeks to employ machine learning methods to determine the regulators of ferroptosis in LUAD. 318 LUAD samples were investigated to determine three ferroptosis molecular phenotypes in LUAD, and then Boruta dimensionality reduction combined with principal component analysis was used to measure the ferroptosis regulation score (FRS) of patients. We additionally presented DeepFerr, a deep learning neural network model, which used the transcriptome map of 11 ferroptosis regulators to predict ferroptosis in LUAD. LASSO, SVM-RFE and elastic net were used to dissect the differential ferroptosis regulators, and the eight pivotal ferroptosis regulators have considerable ferroptosis prediction ability. It was established that RRM2 and AURKA are key suppressors of ferroptosis, and the depletion of RRM2 and AURKA caused an increase in ferroptosis in H358 cells. In addition, not only did they act as pro-proliferative factors that hindered immune infiltration in LUAD, but they were also essential for anti-PD1 therapy and chemotherapy. In summary, this research confirms the regulatory role of RRM2 and AURKA in ferroptosis, and could be useful in predicting individualized treatment for patients with LUAD.