A non-small cell lung cancer-associated circulating nine-microRNA signature for non-invasive lung cancer detection: integrated bioinformatics, machine learning, external validation, and exploratory regulatory network analysis.
5 affiliations
- 1Institute of Laboratory Medicine, Jinling Hospital, School of Life Sciences, Nanjing University, Nanjing, China.
- 2Department of Cardiothoracic Surgery, Jinling Clinical Medical College, Nanjing University of Chinese Medicine, Nanjing, China.
- 3Department of Cardiothoracic Surgery, Jinling Hospital Affiliated to Nanjing Medical University, Nanjing, China.
- 4Department of Cardiothoracic Surgery, Jinling Hospital, Nanjing University, Nanjing, China.
- 5State Key Laboratory of Pharmaceutical Biotechnology, School of Life Sciences, Nanjing University, Nanjing, China.
Abstract
Early and accurate detection of non-small cell lung cancer (NSCLC) remains challenging. Circulating microRNAs (miRNAs) are stable in serum and may provide non-invasive diagnostic biomarkers. This study aimed to identify NSCLC-associated circulating-miRNA signature, evaluate its diagnostic performance across independent sample sets and clinically relevant subgroups, and explore its potential biological associations. Tissue miRNA data from The Cancer Genome Atlas (TCGA) and circulating-miRNA datasets from the Gene Expression Omnibus (GEO) were integrated using tissue-circulation differential-expression screening and machine-learning-based feature selection. GSE188232 served as the primary public development dataset, GSE211692 as a cross-dataset evaluation cohort, and GSE137140 as an additional histologically annotated serum cohort. In GSE137140, the prespecified nine-miRNA feature set was summarized as an outcome-independent composite score and evaluated for overall NSCLC and across histological, age, sex, pathological stage, nodal, and metastatic subgroups. The panel was further assessed by reverse transcription quantitative polymerase chain reaction (RT-qPCR) in an institutional serum cohort of 259 participants. Exploratory regulatory-network, immune-infiltration, pathway, and drug-gene analyses were also performed. Nine circulating miRNAs were retained as the final signature. The nine-miRNA model yielded area under the receiver operating characteristic curve (AUC) values of 0.855 in GSE188232 and 0.991 in GSE211692. In GSE137140, an outcome-independent nine-miRNA composite score based on the prespecified feature set discriminated overall NSCLC (AUC, 0.920), lung adenocarcinoma (LUAD) (AUC, 0.917), and lung squamous cell carcinoma (LUSC) (AUC, 0.937) from non-cancer controls and remained discriminative across age, sex, and stage-specific case-control analyses, including stage I NSCLC (AUC, 0.921), but showed little discrimination between stage I and stage II-IV disease (AUC, 0.513), N0 and N1 + N2 disease (AUC, 0.528), or M0 and M1 disease (AUC, 0.541). In the institutional serum cohort, the panel yielded AUCs of 0.729 for NSCLC vs. combined healthy and benign-disease controls, 0.801 vs. healthy controls, 0.579 vs. benign-disease controls, and 0.744 for stage I NSCLC vs. combined controls. The nine-miRNA signature showed reproducible diagnostic discrimination across public serum datasets and retained diagnostic performance in stage I NSCLC, whereas its ability to distinguish anatomical disease extent was limited. Institutional RT-qPCR evaluation provided additional clinical support, but modest discrimination from benign pulmonary disease and the absence of prospective multicenter validation of a fully locked model currently limit clinical application.
Related
- Identification of candidate selenium-associated transcriptomic biomarkers in chronic obstructive pulmonary disease: an integrated bioinformatics and expression validation study
- Integrative Bioinformatics and Exploratory Experimental Validation of Expression Changes in miR-214-3p, miR-1b, miR-133b-3p, and Directionally Concordant mRNAs in a Rat Model of Acute Myocardial Infarction
- Integrated bioinformatics analysis identifies a ubiquitination-related gene signature in schizophrenia with peripheral clinical validation
- Role of the CXCL8-CXCR2 axis in obesity-related atrial fibrillation: integrated bioinformatics and experimental validation
- cGAS-STING pathway-related genes as candidate molecules associated with cervical intraepithelial neoplasia: an exploratory bioinformatics analysis
BibTeX
@article{wu2026nonsmall,
title = {{A non-small cell lung cancer-associated circulating nine-microRNA signature for non-invasive lung cancer detection: integrated bioinformatics, machine learning, external validation, and exploratory regulatory network analysis.}},
author = {Qiuyue Wu and Penglong Zhao and Qian Wan and Weinan Li and Longxiao Cai and Lan Luo},
journal = {Translational cancer research},
year = {2026},
month = sep,
doi = {10.21037/tcr-2026-1474},
url = {https://europepmc.org/article/MED/42997766},
note = {CC BY-NC-ND},
}
RIS
TY - JOUR
TI - A non-small cell lung cancer-associated circulating nine-microRNA signature for non-invasive lung cancer detection: integrated bioinformatics, machine learning, external validation, and exploratory regulatory network analysis.
AU - Qiuyue Wu
AU - Penglong Zhao
AU - Qian Wan
AU - Weinan Li
AU - Longxiao Cai
AU - Lan Luo
PY - 2026
DA - 2026/09/28
JO - Translational cancer research
DO - 10.21037/tcr-2026-1474
UR - https://europepmc.org/article/MED/42997766
AB - Early and accurate detection of non-small cell lung cancer (NSCLC) remains challenging. Circulating microRNAs (miRNAs) are stable in serum and may provide non-invasive diagnostic biomarkers. This study aimed to identify NSCLC-associated circulating-miRNA signature, evaluate its diagnostic performance across independent sample sets and clinically relevant subgroups, and explore its potential biological associations. Tissue miRNA data from The Cancer Genome Atlas (TCGA) and circulating-miRNA datasets from the Gene Expression Omnibus (GEO) were integrated using tissue-circulation differential-expression screening and machine-learning-based feature selection. GSE188232 served as the primary public development dataset, GSE211692 as a cross-dataset evaluation cohort, and GSE137140 as an additional histologically annotated serum cohort. In GSE137140, the prespecified nine-miRNA feature set was summarized as an outcome-independent composite score and evaluated for overall NSCLC and across histological, age, sex, pathological stage, nodal, and metastatic subgroups. The panel was further assessed by reverse transcription quantitative polymerase chain reaction (RT-qPCR) in an institutional serum cohort of 259 participants. Exploratory regulatory-network, immune-infiltration, pathway, and drug-gene analyses were also performed. Nine circulating miRNAs were retained as the final signature. The nine-miRNA model yielded area under the receiver operating characteristic curve (AUC) values of 0.855 in GSE188232 and 0.991 in GSE211692. In GSE137140, an outcome-independent nine-miRNA composite score based on the prespecified feature set discriminated overall NSCLC (AUC, 0.920), lung adenocarcinoma (LUAD) (AUC, 0.917), and lung squamous cell carcinoma (LUSC) (AUC, 0.937) from non-cancer controls and remained discriminative across age, sex, and stage-specific case-control analyses, including stage I NSCLC (AUC, 0.921), but showed little discrimination between stage I and stage II-IV disease (AUC, 0.513), N0 and N1 + N2 disease (AUC, 0.528), or M0 and M1 disease (AUC, 0.541). In the institutional serum cohort, the panel yielded AUCs of 0.729 for NSCLC vs. combined healthy and benign-disease controls, 0.801 vs. healthy controls, 0.579 vs. benign-disease controls, and 0.744 for stage I NSCLC vs. combined controls. The nine-miRNA signature showed reproducible diagnostic discrimination across public serum datasets and retained diagnostic performance in stage I NSCLC, whereas its ability to distinguish anatomical disease extent was limited. Institutional RT-qPCR evaluation provided additional clinical support, but modest discrimination from benign pulmonary disease and the absence of prospective multicenter validation of a fully locked model currently limit clinical application.
SN - 2218-676X
SN - 2219-6803
ER -