Abstract
Anoikis dysregulation and fatty acid metabolism reprogramming are critical for colorectal cancer (CRC) progression, but their integrated role in CRC prognosis remains unclear. We used transcriptomic and clinical data of CRC patients from the cancer genome atlas and gene expression omnibus (GSE12945) databases. Differential analysis, NMF clustering, univariate Cox, and LASSO regression were used to identify (Anoikis-fatty acid metabolism-related gene) AFRGs and construct a prognostic model. The model's performance was validated via receiver operating characteristic curves, nomogram, and immune infiltration analysis. We identified 232 differentially expressed AFRGs and 8 prognostic AFRGs (NMF clustering divided patients into 2 clusters with distinct survival). A 5-gene (BRCA1, CD36, ENO3, INHBB, PHLDA2) model showed robust predictive efficacy (1-, 3-, 5-year OS area under the curve: 0.889/0.795/0.740 in GSE12945). High-risk patients had higher PDCD1/CTLA4 expression and CD8⁺ T cell infiltration. The nomogram (risk score + clinicopathological factors) had high accuracy (area under the curve: 0.788/0.807/0.833). This AFRG-based model is a reliable prognostic tool for CRC and provides insights for personalized treatment. Limitations include retrospective data and small gene expression omnibus cohort, requiring prospective validation.