An Intelligent Classification System for Cancer Detection Based on DNA Methylation Using ML and Semantic Knowledge in...

Area of Science:

• Biomedical Engineering
• Computational Biology
• Machine Learning in Healthcare

Background:

• Wearable sensing devices are increasingly used for assessing patient wellness and diagnosing chronic conditions.
• Social network analysis (SNA) and machine learning (ML) are emerging tools in healthcare for analyzing complex data.
• DNA methylation, an epigenetic process, is linked to various health issues, including cancer.

Purpose of the Study:

• To develop an intelligent system using wearable sensors for cancer classification based on DNA methylation data.
• To address challenges of class imbalance and high dimensionality in large biomedical datasets like The Cancer Genome Atlas (TCGA).
• To enhance the accuracy and sensitivity of cancer detection, particularly for minority classes.

Main Methods:

• A mixed-sampling imbalanced data ensemble classification technique was developed, incorporating Intelligent Synthetic Minority Oversampling (SMOTE) and Tomek Link methods.
• The system utilizes wearable biomedical sensors to collect DNA methylation data.
• A deep forest (GC-Forest) algorithm with cascading forest structures was employed for final classification.

Main Results:

• The proposed technique effectively handles class imbalance and noise in large datasets.
• The method demonstrated increased sensitivity for the minority class in cancer classification.
• Classification accuracy for the majority class was maintained while improving overall detection capabilities.

Conclusions:

• The developed mixed-sampling imbalanced data ensemble classification technique shows promise for accurate cancer detection using DNA methylation data from wearable sensors.
• This approach offers a viable solution for improving diagnostic accuracy in complex, imbalanced biomedical datasets.
• The study highlights the potential of integrating wearable technology, epigenetic data, and advanced machine learning for personalized cancer diagnostics.