Machine Learning Reveals Impacts of Smoking on Gene Profiles of Different Cell Types in Lung

Area of Science:

• Oncology
• Genomics
• Computational Biology

Background:

• Smoking is a major risk factor for lung diseases like COPD and lung cancer.
• Harmful chemicals in tobacco smoke damage lung tissue and function.
• Limited research exists on smoking's impact on gene expression in specific lung cell types.

Purpose of the Study:

• To analyze single-cell gene expression profiles in various lung cell types.
• To identify potential gene markers associated with smoking status.
• To understand the molecular mechanisms of smoking-induced lung damage.

Main Methods:

• Single-cell RNA sequencing of lung cells from active, former, and never smokers.
• Machine learning techniques, including incremental feature selection.
• Analysis of gene expression data across lung endothelial, epithelial, immune, and stromal cells.

Main Results:

• Identified smoking-associated genes in different lung cell types: B2M, EEF1A1, TPT1 (endothelial); FTL, MT-ATP8 (epithelial); HLA-B, HLA-C (immune); HSP90B1, LCN2 (stroma).
• Developed quantitative rules to represent smoking-related gene expression patterns.
• Demonstrated the utility of machine learning in identifying molecular markers of smoking.

Conclusions:

• Machine learning can effectively identify gene expression signatures related to smoking in specific lung cell populations.
• The identified genes serve as potential biomarkers for smoking status and lung damage.
• This research advances the molecular understanding of smoking's effects on the lung and supports future cancer research.