Comprehensive Serum Glycopeptide Spectra Analysis Combined with Machine Learning for Early Detection of Lung Cancer:...

Area of Science:

• Biomarker Discovery
• Proteomics
• Machine Learning in Diagnostics

Background:

• Lung cancer is a leading cause of cancer death worldwide.
• Conventional screening methods like CT scans have limitations including cost and radiation exposure.
• Non-invasive blood tests are needed to improve screening participation and early detection.

Purpose of the Study:

• To develop and validate a blood-based diagnostic model for lung cancer detection.
• To assess the utility of enriched glycopeptides (EGPs) and conventional tumor markers for early lung cancer diagnosis.

Main Methods:

• Analyzed serum samples from 199 lung cancer patients and 590 healthy controls.
• Measured nine conventional tumor markers and 1688 EGPs using liquid chromatography-mass spectrometry.
• Utilized machine learning models to build a predictive diagnostic algorithm.

Main Results:

• Identified specific glycopeptides (AT271-FSG and MG70-FSG) that distinguish lung cancer patients from healthy individuals.
• The Comprehensive Serum Glycopeptide Spectra Analysis (CSGSA) model achieved an ROC-AUC of 0.935 for overall detection.
• The model demonstrated high accuracy for early-stage (Stage I) lung cancer detection with an ROC-AUC of 0.914.

Conclusions:

• A novel blood-based diagnostic approach combining EGPs and machine learning significantly improves lung cancer detection accuracy.
• This method holds potential for early lung cancer diagnosis, which is crucial for improving patient outcomes.
• The developed CSGSA model represents a significant advancement in non-invasive cancer diagnostics.