Abstract
BACKGROUND: Cigarette smoking is associated with an increased risk of developing respiratory diseases and various types of cancer. Early identification of such unfavorable outcomes in patients who smoke is critical for optimizing personalized medical care. METHODS: Here, we perform a comprehensive analysis using Systems Biology tools of publicly available data from a total of 6 transcriptomic studies, which examined different specimens of lung tissue and/or cells of smokers and nonsmokers to identify potential markers associated with lung cancer. RESULTS: Expression level of 22 genes was capable of classifying smokers from non-smokers. A machine learning algorithm revealed that AKR1B10 was the most informative gene among the 22 differentially expressed genes (DEGs) accounting for the classification of the clinical groups. AKR1B10 expression was higher in smokers compared to non-smokers in datasets examining small and large airway epithelia, but not in the data from a study of sorted alveolar macrophages. Moreover, AKR1B10 expression was relatively higher in lung cancer specimens compared to matched healthy tissue obtained from nonsmoking individuals. Although the overall accuracy of AKR1B10 expression level in distinction between cancer and healthy lung tissue was 76%, with a specificity of 98%, our results indicated that such marker exhibited low sensitivity, hampering its use for cancer screening such specific setting. CONCLUSION: The systematic analysis of transcriptomic studies performed here revealed a potential critical link between AKR1B10 expression, smoking and occurrence of lung cancer.