# Enhancing colorectal cancer diagnosis with the ferroptosis marker GPX4 and serum biomarkers: A retrospective analysis and machine learning approach

## TL;DR

This study explores using GPX4, a ferroptosis marker, and machine learning to improve early diagnosis of colorectal cancer.

## Contribution

The novel integration of GPX4 with traditional biomarkers and machine learning improves CRC diagnostic accuracy.

## Key findings

- Low GPX4 expression correlates with higher survival rates in CRC patients.
- Combining GPX4 with CEA, CA19-9, and serum iron improves diagnostic accuracy.
- Machine learning models using these markers achieved high AUC values.

## Abstract

Colorectal cancer (CRC) remains a significant global health burden characterized by high incidence and mortality. Although early diagnosis is crucial for improving prognosis, current screening methods are limited by insufficient sensitivity. Ferroptosis, an iron-dependent form of regulated cell death, is implicated in tumor progression, with Glutathione peroxidase 4 (GPX4) serving as a key regulator in this process.

This study aims to evaluate the potential of GPX4 as an early diagnostic biomarker for CRC and to construct an integrated predictive model combining ferroptosis-related indicators with machine learning algorithms to enhance diagnostic precision.

Clinical data and GPX4 gene expression profiles were retrieved from the TCGA-COAD database for survival, differential expression, single-cell transcriptomic, and immune infiltration analyses. Subsequently, GPX4 expression was validated in CRC tissues, adjacent normal tissues, and cell lines with varying metastatic potentials using immunohistochemistry (IHC), qRT-PCR, and Western blotting. Serum levels of carcinoembryonic antigen (CEA), carbohydrate antigen 19-9 (CA19-9), GPX4, and serum iron were quantified in 120 CRC patients and 120 healthy controls. The diagnostic efficacy of individual markers was assessed using receiver operating characteristic (ROC) curves and Kappa analysis. Finally, nine machine learning algorithms were employed to develop a combinatorial diagnostic model based on serum ferroptosis-related indicators, with performance evaluated via ROC and decision curve analysis (DCA).

Ferroptosis plays a critical role in CRC pathogenesis and progression. GPX4 was identified as a robust biomarker for early diagnosis. The integrated machine learning model, incorporating GPX4, CEA, CA19-9, and serum iron, demonstrated superior diagnostic performance compared to conventional markers, offering a promising strategy for the early detection of CRC.

This study has made significant contributions to exploring diagnostic methods for colorectal cancer, highlighting the following key points:•Critical Role of GPX4 in Colorectal Cancer Survival Status and Immune Cell Infiltration:●It was found that low GPX4 expression was associated with a higher survival rate in colorectal cancer patients, providing new biomarker clues for diagnosis and treatment.●The correlation between GPX4 and various immune cells was demonstrated, and immunohistochemical analysis of cancer tissues and surrounding tissues revealed the relationship between GPX4, tumor immune microenvironment, and tumor parenchyma.•Innovations in Colorectal Cancer Diagnosis by Combining GPX4 Serum Markers:●The combination of GPX4 with traditional serum markers such as CEA and CA199 significantly improved the diagnostic accuracy of colorectal cancer, offering new insights for clinical diagnosis.●An analysis of combined detection of ferroptosis markers and traditional serum markers presented a novel and more accurate method for colorectal cancer diagnosis.•Cutting-Edge Application of Machine Learning in Colorectal Cancer Diagnosis:●Utilizing various machine learning algorithms, diagnostic models for colorectal cancer were constructed, achieving high AUC values and demonstrating good predictive performance.●Through machine learning methods, new approaches and possibilities for personalized diagnosis of colorectal cancer were provided, showcasing the potential of technology in the medical field.

Critical Role of GPX4 in Colorectal Cancer Survival Status and Immune Cell Infiltration:●It was found that low GPX4 expression was associated with a higher survival rate in colorectal cancer patients, providing new biomarker clues for diagnosis and treatment.●The correlation between GPX4 and various immune cells was demonstrated, and immunohistochemical analysis of cancer tissues and surrounding tissues revealed the relationship between GPX4, tumor immune microenvironment, and tumor parenchyma.

It was found that low GPX4 expression was associated with a higher survival rate in colorectal cancer patients, providing new biomarker clues for diagnosis and treatment.

The correlation between GPX4 and various immune cells was demonstrated, and immunohistochemical analysis of cancer tissues and surrounding tissues revealed the relationship between GPX4, tumor immune microenvironment, and tumor parenchyma.

Innovations in Colorectal Cancer Diagnosis by Combining GPX4 Serum Markers:●The combination of GPX4 with traditional serum markers such as CEA and CA199 significantly improved the diagnostic accuracy of colorectal cancer, offering new insights for clinical diagnosis.●An analysis of combined detection of ferroptosis markers and traditional serum markers presented a novel and more accurate method for colorectal cancer diagnosis.

The combination of GPX4 with traditional serum markers such as CEA and CA199 significantly improved the diagnostic accuracy of colorectal cancer, offering new insights for clinical diagnosis.

An analysis of combined detection of ferroptosis markers and traditional serum markers presented a novel and more accurate method for colorectal cancer diagnosis.

Cutting-Edge Application of Machine Learning in Colorectal Cancer Diagnosis:●Utilizing various machine learning algorithms, diagnostic models for colorectal cancer were constructed, achieving high AUC values and demonstrating good predictive performance.●Through machine learning methods, new approaches and possibilities for personalized diagnosis of colorectal cancer were provided, showcasing the potential of technology in the medical field.

Utilizing various machine learning algorithms, diagnostic models for colorectal cancer were constructed, achieving high AUC values and demonstrating good predictive performance.

Through machine learning methods, new approaches and possibilities for personalized diagnosis of colorectal cancer were provided, showcasing the potential of technology in the medical field.

These innovative findings not only expand the research scope of colorectal cancer diagnosis but also offer new directions and strategies for early diagnosis and treatment in future clinical practice.