Accumulated evidence indicates that various types of
miRNA are aberrantly expressed in
lung cancer and secreted into the bloodstream. For this study, we constructed a serum diagnostic classifier based on detailed bioinformatics analysis of
miRNA profiles from a training cohort of 143
lung adenocarcinoma patients and 49 healthy subjects, resulting in a 20
miRNA-based classifier. Validation performed with an independent cohort of samples from
lung adenocarcinoma patients (n = 110), healthy subjects (n = 52), and benign
pulmonary disease patients (n = 47) showed a sensitivity of 89.1% and specificity of 94.9%, with an area under the curve value of 0.958. Notably, 90.8% of Stage I
lung adenocarcinoma cases were correctly diagnosed. Interestingly, this classifier also detected squamous and large cell lung
carcinoma cases at relatively high rates (70.4% and 70.0%, respectively), which appears to be consistent with organ site-dependent
miRNA expression in
cancer tissues. In contrast, we observed significantly lower rates (0-35%) using samples from 96 cases of
cancer in other major organs, with
breast cancer the lowest. These findings warrant a future study to realize its clinical application as a part of diagnostic procedures for
lung cancers, for which early detection and surgical removal is presently the only hope for eventual cure.