Smoking Classification Using Novel Plasma Cytokines by implementing Machine Learning and Statistical Methods.

Area of Science:

• Biomarkers and Disease Classification
• Computational Biology and Bioinformatics

Background:

• Smoking is a leading cause of preventable death, linked to numerous diseases like COPD, cardiovascular disease, cancer, and diabetes.
• Cytokines, as inflammatory biomarkers, are mechanistically associated with smoking and its associated health risks.

Purpose of the Study:

• To apply machine learning algorithms for quantitative assessment of cytokine contributions to smoking-related diseases.
• To classify smoking status using plasma cytokine profiles and identify key biomarkers.

Main Methods:

• Utilized k Nearest Neighbor (k-NN) and Random Forest machine learning algorithms on 63 plasma cytokines.
• Employed k-fold cross-validation and hyperparameter tuning for performance optimization.
• Evaluated model performance using Area Under the Receiver Operating Characteristic (AUROC) curves.

Main Results:

• k-NN achieved an AUROC of 0.87 (95% CI: 0.823–0.917).
• Random Forest demonstrated superior performance with a perfect AUROC of 1.0 (95% CI: 1–1).
• Identified common significant cytokines: LIF, IL22, G-CSF/CSF-3, and TRAIL, crucial for classification.

Conclusions:

• Machine learning effectively classifies smoking status based on plasma cytokine profiles.
• Identified cytokines can serve as biomarkers for smoking-related diseases, facilitating precision medicine.
• The study highlights the potential of translating molecular findings into clinical practice for targeted interventions.