Identification of PIWI-interacting RNAs based models for lung adenocarcinoma early detection: a multicenter cohort study

Abstract Early detection of lung adenocarcinoma (LUAD) remains a major clinical challenge despite the widespread application of low-dose computed tomography (LDCT). Circulating PIWI-interacting RNAs (piRNAs), characterized by tumor-specific expression and high stability, offer promise as non-invasive biomarkers. To improve the diagnostic precision of LDCT screening, we performed a large multi-center study integrating paired tissue–serum omics profiling with machine learning–based biomarker discovery. From 1,521 serum samples (1,033 LUAD, 89 benign pulmonary nodules, and 399 healthy controls), two tumor-derived PIWI-interacting RNAs (piR-hsa-8393202 and piR-hsa-8429916) were identified as highly stable, LUAD-specific molecules closely associated with disease progression. A 2-piRNA diagnostic signature demonstrated robust performance for early-stage LUAD (training AUC = 0.918; validation AUC = 0.863) and adenocarcinoma in situ (training AUC = 0.902; validation AUC = 0.907). Notably, when applied to LDCT-detected indeterminate pulmonary nodules, this signature significantly improved malignant nodule identification (AUC = 0.883), outperforming conventional serum biomarkers such as carcinoembryonic antigen and cytokeratin 19 fragment antigen 21–1. Functional assays further revealed that these piRNAs promote tumor cell proliferation and suppress apoptosis, supporting their oncogenic activity. Collectively, this study establishes circulating piRNAs as non-invasive and mechanistically relevant biomarkers for molecular stratification of pulmonary nodules within LDCT screening programs, providing a clinically applicable tool to refine early lung cancer diagnosis and guide individualized management.