Integrative approach for classifying male tumors based on DNA methylation 450K data

Area of Science:

• Oncology
• Bioinformatics
• Computational Biology

Background:

• Malignancies like bladder urothelial carcinoma, colon adenocarcinoma, liver hepatocellular carcinoma, lung adenocarcinoma, and prostate adenocarcinoma significantly affect male health.
• Accurate cancer classification is crucial for effective treatment strategies and improved patient outcomes.

Purpose of the Study:

• To introduce an innovative method for enhanced male tumor classification using gene selection from high-dimensional DNA methylation data.
• To assess the reliability of DNA methylation data in distinguishing five prevalent male cancers from normal tissues.

Main Methods:

• Utilized DNA methylation 450K data from The Cancer Genome Atlas (TCGA) database.
• Employed chi-square test for dimensionality reduction and L1 penalized logistic regression for feature selection.
• Applied a stacking ensemble learning technique integrating seven common multiclassification models.

Main Results:

• The proposed ensemble learning model significantly outperformed individual base classification models.
• Achieved an overall accuracy of 99.2% on independent testing data for classifying five male cancer types.
• Demonstrated the effectiveness of DNA methylation data and ensemble learning in cancer classification.

Conclusions:

• The developed method offers a highly accurate approach for classifying common male tumors.
• This study provides novel insights for the early detection and treatment of male-specific cancers.
• Highlights the potential of integrating gene selection and ensemble learning in cancer diagnostics.