Exploring Volatile Organic Compounds in Breath for High-Accuracy Prediction of Lung Cancer

Area of Science:

• Biochemistry
• Computational Biology
• Medical Diagnostics

Background:

• Lung cancer diagnosis relies on imaging, which has limitations like low accuracy (X-rays) or radiation exposure and cost (CT scans).
• Breathomics, analyzing volatile organic compounds (VOCs) in breath, presents a promising noninvasive strategy for disease biomarker discovery.
• VOCs can reflect metabolic alterations associated with diseases, potentially serving as early lung cancer indicators.

Purpose of the Study:

• To develop and validate a machine learning model for lung cancer prediction using exhaled volatile organic compounds (VOCs).
• To assess the diagnostic performance of breathomics compared to conventional lung cancer screening methods.

Main Methods:

• Quantitative analysis of 116 volatile organic compounds (VOCs) in breath samples from lung cancer patients and healthy controls using Selected Ion Flow Tube Mass Spectrometry (SIFT-MS).
• Development of a predictive model using eXtreme Gradient Boosting (XGBoost) machine learning algorithm based on VOC measurements.
• Adjustment for environmental VOCs to refine the accuracy of the lung cancer prediction model.

Main Results:

• The XGBoost model achieved high performance, with an initial accuracy of 0.89, sensitivity of 0.82, specificity of 0.94, and AUC of 0.95.
• After adjusting for environmental VOCs, the model's performance improved to 0.92 accuracy, 0.96 sensitivity, 0.88 specificity, and 0.98 AUC.
• The developed breathomics approach demonstrated superior predictive capability compared to existing methods.

Conclusions:

• A quantitative VOCs database combined with an XGBoost classifier offers a robust platform for noninvasive lung cancer prediction.
• Breathomics represents a significant advancement in early lung cancer detection, potentially improving patient outcomes.
• This approach provides a cost-effective and noninvasive alternative for lung cancer screening.