Background: Lymph node (LN) status is vital to indicate and evaluate the curative potential of relatively early gastric cancer (GC; T1–T2) treatment (endoscopic or surgery). Currently, there is a lack of robust and convenient methods to identify such metastasis before therapeutic decision-making; therefore, there is an urgent need to identify biomarkers that could aid the identification of patients with LN metastasis.
Methods: Genome-wide expression profiles of long noncoding RNA (lncRNA) in primary T1 gastric cancer data from The Cancer Genome Atlas (TCGA) was used to identify an lncRNA‑expression signature capable of detecting LN metastasis of GC, and establish a 10-lncRNA risk‑prediction model based on deap learning. The performance of the lncRNA panel in diagnosing LN metastasis was evaluated using both in silico and clinical validation methods. In silico validation was conducted using TCGA and Asian Cancer Research Group (ACRG) datasets. Clinical validation was performed on T1 and T2 patients, and the panel's efficacy was compared with that of traditional tumor markers and computed tomography (CT) scans.
Results: Profiling of genome-wide RNA expression identified a panel of lncRNA to predict LN metastasis in T1 stage gastric cancer (area under the curve (AUC) = 0.961). A 10-lncRNA risk-prediction model was then constructed, which was validated successfully in T1 and T2 datasets (TCGA, AUC = 0.852; ACRG, AUC = 0.834). Thereafter, the clinical performance of the lncRNA panel was validated in clinical cohorts (T1, AUC = 0.812; T2, AUC = 0.805; T1+T2, AUC = 0.764). Notably, the 10-lncRNA panel demonstrated significantly better performance compared with CT and conventional tumor markers (carcinoembryonic antigen and carbohydrate antigen 19-9).
Conclusions: The novel 10-lncRNA could diagnose LN metastasis robustly in relatively early gastric cancer (T1–T2), with promising clinical potential.