The accurate selection of serum metabolomic biomarkers for early lung cancer screening remains a significant challenge in the clinical context. Consequently, this study introduces the Red Fox Optimization (RFO) that integrates Dynamic Mutation Late Acceptance Hill Climbing (DM-LAHC), with the aim of selecting a panel of serum metabolomic biomarkers suitable for distinguishing between benign and malignant pulmonary nodules. The key innovation is the dynamic adjustment of the mutation probability in the Late Acceptance Hill Climbing algorithm, which greatly enhances the local search capabilities. And the RFO's reproduction mechanism has been improved through the utilization of a more efficient interpolation form. The biomarker selection model employs a multi-objective fitness function that takes into account both accuracy and quantity. After this, the optimal model yielded a biomarker panel, including Inosine, Hippuric acid, Alanine, and other metabolites. This model demonstrates outstanding performance on an independent test dataset, achieving a fitness value of 0.9136, an AUC (Area Under the Curve) of 0.9926, a sensitivity of 0.9643, and a specificity of 0.9412. Furthermore, the clinical net benefit is highlighted across various risk thresholds by decision curve analysis. These results underscore the significance of DM-LAHC aided RFO in the selection of serum metabolomic biomarkers for lung cancer. The supporting source codes of this work can be found at: https://github.com/zzl2022/DM-LAHC-aided-RFO.