Colorectal cancer (CRC) is strikingly immunologically heterogeneous but the mechanistic basis for the immune-refractory “cold” feature is unknown. On basis of The Cancer Genome Atlas-Colon Adenocarcinoma (TCGA-COAD) data, we distinguished tumors into “cold” and “hot” subtypes. We discovered peptide deformylase (PDF) as a marker of cold CRC by multi-algorithm differential expression and machine learning (Support Vector Machine (SVM), Random Forest (RF), XGBoost). High expression of PDF correlated with poor prognosis, low immune infiltration and high level of oxidative phosphorylation (OXPHOS). Mechanismwise, according to extant literature evidence and bioinformatic inference, PDF is speculated to deformylate N-terminal of mitochondrial new-peptides, thereby contributing to maturation of ETC and oxidative phosphorylation to afford ATP supplies to accelerate tumor growth.Concurrently, this process may curtail the release of N-formyl peptides—damage-associated molecular patterns that recruit CD8⁺ T cells and macrophages via Formyl Peptide Receptor 1 (FPR1)—thereby reinforcing an immune-excluded microenvironment. Bioinformatic analyses further nominated Poly(rC)-binding protein 1 (PCBP1) as a potential transcriptional regulator of PDF, a finding supported by an independent Clinical Proteomic Tumor Analysis Consortium—Phase 2 (CPTAC-2) cohort. Our work suggests a PCBP1–PDF–OXPHOS axis associated with immune suppression in CRC.