Transcriptome-Based Six-Gene Fatty Acid Metabolism Signature for Prognosis and Predicted Immunotherapy Response in Lung Adenocarcinoma: Cross-Population Validation

Abstract

Objectives: Lung adenocarcinoma (LUAD) is molecularly heterogeneous, and the prognostic relevance of fatty acid metabolism (FAM) remains incompletely defined. We aimed to develop a concise FAM-associated prognostic signature and examine its associations with the immune microenvironment and candidate therapeutic vulnerabilities. Methods: TCGA-LUAD transcriptomic and survival data were integrated with MSigDB FAM gene sets. Univariate Cox and elastic-net Cox regression were used to derive a risk score. The locked formula was evaluated in a Japanese cohort (GSE31210) and a U.S. cohort (GSE72094). Immune-infiltration algorithms as well as TIDE, GDSC2, and CPTAC data were used for exploratory immune, drug sensitivity, and protein-level analyses. Results: The six-gene signature comprised CYP4B1, ACOXL, DPEP2, HPGDS, CA4, and ALOX15. High-risk patients had shorter overall survival in the TCGA and both external cohorts (GSE31210, log-rank p = 0.0039; GSE72094, p < 0.0001). The risk score remained independently prognostic after adjustment for age, sex, and clinical stage. High-risk tumours showed lower immune and stromal signals, greater immune exclusion, and a lower TIDE-predicted ICB response proportion. GDSC2 analyses and expression comparisons identified associations with predicted drug sensitivity and lipogenic target expression. Five detectable signature proteins were less abundant in tumours in CPTAC data. Conclusions: This signature stratified patients by prognosis in two geographically distinct external cohorts and generated testable metabolic and immune hypotheses. Prospective validation, assay standardisation, and functional studies are required before clinical use.