Colon cancer diagnosis and staging classification based on machine learning and bioinformatics analysis

Area of Science:

• Oncology
• Bioinformatics
• Genomics

Background:

• Advanced colon cancer metastasis presents significant treatment challenges.
• Early diagnosis and staging of colon cancer are crucial for improving patient prognosis.
• Gene expression profiling offers potential for identifying diagnostic and prognostic markers.

Purpose of the Study:

• To identify key gene expression markers for colon cancer diagnosis and staging.
• To develop machine learning models for accurate colon cancer detection and classification.
• To discover novel genes associated with colon cancer prognosis.

Main Methods:

• Weighted Gene Co-expression Network Analysis (WGCNA) to identify cancer-correlated gene modules.
• Least Absolute Shrinkage and Selection Operator (Lasso) algorithm for feature gene extraction.
• Random Forest (RF), Support Vector Machine (SVM), and decision trees for classification; Protein-Protein Interaction (PPI) network analysis for survival analysis.

Main Results:

• RF model achieved high accuracy (99.81%) in distinguishing colon cancer from healthy controls.
• RF model demonstrated notable accuracy (91.5%) in diagnosing colon cancer stages I-IV.
• Eight genes (GCNT2, GLDN, SULT1B1, UGT2B15, PTGDR2, GPR15, BMP5, CPT2) were identified as significant prognostic markers.

Conclusions:

• Gene expression analysis combined with machine learning effectively diagnoses colon cancer and its stages.
• The identified prognostic genes hold potential for targeted therapeutic strategies and improved patient outcomes.
• This study provides a foundation for developing novel diagnostic and prognostic tools for colon cancer.