The emerging role of machine learning-based methods in cancer classification using microRNA

Area of Science:

• Biomarkers and Diagnostics
• Computational Biology
• Oncology

Background:

• Accurate cancer detection and classification are vital for patient outcomes but challenging with conventional methods.
• MicroRNAs (miRNAs) show promise as stable biomarkers in bodily fluids for non-invasive cancer diagnostics.
• miRNAs play roles as oncogenes or tumor suppressors in cancer progression.

Purpose of the Study:

• To review the application of machine learning (ML) models with microRNA (miRNA) data for cancer diagnostics.
• To highlight the potential of miRNA-driven ML in differentiating tumor types and subtypes.
• To explore the role of ML and miRNAs in advancing personalized cancer treatment.

Main Methods:

• Utilizing machine learning algorithms (e.g., Random Forest, Support Vector Machines, deep learning) to analyze miRNA expression data.
• Employing feature engineering and selection techniques, including recursive ensemble selection and miRNA-mRNA network analysis.
• Validating miRNA signatures in bodily fluids (blood, urine, saliva, feces) for various cancer types.

Main Results:

• ML models effectively identify discriminative miRNAs for classifying cancers like breast, lung, colorectal, and kidney cancer.
• Integration of ML and miRNA data significantly improves the differentiation between cancerous and normal tissues.
• Specific miRNA signatures show high accuracy in diagnosing colorectal cancer and classifying breast cancer subtypes.

Conclusions:

• miRNA-driven ML models offer a powerful, non-invasive approach for accurate cancer diagnostics.
• These models enhance the identification of clinically relevant biomarkers for personalized medicine.
• The integration of ML and miRNA analysis holds transformative potential for cancer care.