Machine Learning-Enabled Non-Invasive Screening of Tumor-Associated Circulating Transcripts for Early Detection of...

Area of Science:

• Oncology
• Biomarker Discovery
• Computational Biology

Background:

• Colorectal cancer (CRC) remains a leading cause of cancer mortality globally.
• Accurate, non-invasive diagnostic methods are crucial for early CRC detection and improved patient survival.
• Tumor-associated circulating transcripts (TACTs) are emerging as potential biomarkers for cancer detection.

Purpose of the Study:

• To evaluate the diagnostic utility of TACTs for colorectal cancer (CRC) detection.
• To integrate TACT markers into machine learning (ML) models to enhance diagnostic accuracy.
• To identify the optimal ML model for CRC detection using TACT biomarkers.

Main Methods:

• The study assessed five machine learning models: Generalized Linear Model, Random Forest, Gradient Boosting Machine, Deep Neural Network (DNN), and AutoML.
• TACTs were utilized as biomarkers within these models for CRC detection.
• Model performance was evaluated based on sensitivity and specificity.

Main Results:

• The Deep Neural Network (DNN) model demonstrated optimal performance.
• The DNN model achieved a sensitivity of 85.7% and a specificity of 90.9% for CRC detection.
• High diagnostic performance was particularly noted for early-stage colorectal cancer.

Conclusions:

• Combining TACT markers with AI-based analysis presents a scalable and precise approach for non-invasive CRC screening.
• This integrated approach significantly advances non-invasive cancer diagnostics.
• The findings support the potential of TACTs and AI for improving early CRC detection and patient outcomes.