Identification PDF as a biomarker of “cold” CRC using integrated bioinformatics and machine learning

A recent study demonstrates that integrating multi-algorithm bioinformatics with machine learning models reveals peptide deformylase (PDF) as a novel biomarker for “cold” Colorectal cancer. High expression of PDF correlated with poor prognosis, low immune infiltration and high level of oxidative phosphorylation (OXPHOS) in Colorectal cancer. This may offer a metabolic–immune roadmap for precision immunotherapy of “cold” Colorectal cancer.

Colorectal cancer (CRC) exhibits profound immune heterogeneity, yet the molecular underpinnings of immunotherapy-resistant “cold” tumors remain elusive. In a recent study published in Advanced Cancer Research, Chen et al. combined TCGA-COAD bioinformatics with machine learning frameworks—including Support Vector Machines, Random Forest, and XGBoost—to systematically dissect the molecular distinctions between immune-hot and immune-cold CRC subtypes. By coupling multi-algorithm differential analysis with independent validation using the CPTAC-2 cohort, the authors identified peptide deformylase (PDF) as a robust candidate biomarker of the cold phenotype. High PDF expression correlates with poor prognosis, diminished immune cell infiltration, and a metabolic shift toward oxidative phosphorylation. Mechanistically, the study delineates a PCBP1–PDF–OXPHOS regulatory axis​ that not only fuels tumor bioenergetics but also suppresses N-formyl peptide–mediated recruitment of CD8⁺ T cells and macrophages, thereby reinforcing an immunosuppressive microenvironment. This work exemplifies how AI-assisted multi-omics integration can bridge the gap between static genomic data and dynamic functional phenotypes, uncovering clinically relevant targets for overcoming immune resistance in precision oncology.

Key highlights include:

• Machine learning-driven subtyping: Integration of SVM, Random Forest, and XGBoost models enables robust stratification of immune-hot versus cold colorectal tumors.
• Novel biomarker discovery:  Peptide deformylase (PDF) is identified as a reliable protein-level indicator of the cold tumor phenotype and poor clinical outcomes.
• Mechanistic insight: The proposed PCBP1–PDF–OXPHOS axis links metabolic reprogramming to immune exclusion, providing a functional explanation for immunotherapy resistance.
• Cross-cohort validation: Independent verification using the CPTAC-2 proteomics dataset strengthens the translational relevance of the findings.
• Therapeutic implication: Targeting the PDF-associated metabolic–immune crosstalk offers a potential strategy to reprogram cold tumors and improve response to immunotherapy.

Citation: Chen H, Zhong Z, Lin J, Shi X. Identification PDF as a biomarker of “cold” CRC using integrated bioinformatics and machine learning. Adv. Cancer Res. 2026(3):0013, https://doi.org/10.55092/acr20260013.