Development and validation of a machine learning-based prognostic model using mitochondrial dysfunction-related genes
Chang Liu1, Yaxuan Li1, Jiaxu Song1
1First School of Clinical Medicine, Nanjing Medical University, Nanjing, China.
Translational Cancer Research
|November 14, 2025
Summary
This study developed a machine learning model using mitochondria-related genes to predict colorectal cancer (CRC) patient outcomes. The 7-gene signature effectively stratifies risk and guides personalized therapy.
Area of Science:
- Oncology
- Genomics
- Bioinformatics
Background:
- Colorectal cancer (CRC) is a leading cause of cancer mortality, with a need for novel prognostic biomarkers.
- Mitochondrial dysfunction is a cancer hallmark, influencing metabolism and immune evasion, but its prognostic role in CRC is underexplored.
- Current prognostic models for CRC lack deep insights into mitochondrial dysfunction.
Purpose of the Study:
- To develop a machine learning (ML)-based prognostic model utilizing mitochondria-related genes (MRGs).
- To stratify colorectal cancer patients based on their risk.
- To guide personalized therapeutic strategies for CRC.
Main Methods:
- Analysis of RNA sequencing and clinical data from The Cancer Genome Atlas Program and Gene Expression Omnibus datasets.
- Identification of 316 differentially expressed MRGs and key pathways via functional enrichment analysis.
- Development and validation of a 7-gene risk model using ML algorithms for overall survival (OS) prediction, immune infiltration, drug sensitivity, and immunotherapy response assessment.
Main Results:
- A 7-gene signature (TPM2, GSTM1, CYP11A1, SCN4A, LEP, PPARGC1A, NRG1) effectively stratified CRC patients into high- and low-risk groups (P<0.05).
- High-risk scores correlated with poorer overall survival.
- Significant differences were observed between risk groups in immune cell infiltration, immune cell function, and drug sensitivity.
Conclusions:
- The ML-based model using MRGs accurately predicts CRC prognosis, immune microenvironment, and therapeutic response.
- The 7-gene signature provides a framework for precision oncology by guiding risk stratification and targeted therapy.
- This approach bridges mitochondrial biology with clinical outcomes in colorectal cancer management.
