A decade-ahead signal of lung cancer from circulating exosomal sncRNAs
Presenter: Zhuokun Feng, MD Session: Diagnostic Biomarkers 2 Time: 4/21/2026 2:00:00 PM → 4/21/2026 5:00:00 PM
Authors
Zhuokun Feng 1 , Masaki Nasu 2 , Lauren Higa 2 , Isam M. Ibrahim 2 , Loïc L. Marchand 3 , Youping Deng 1 1 Department of Quantitative Health Sciences, John A. Burns School of Medicine, University of Hawaii at Manoa, Honolulu, HI, 2 John A. Burns School of Medicine, University of Hawaii at Manoa, Honolulu, HI, 3 University of Hawaii Cancer Center, University of Hawaii at Manoa, Honolulu, HI
Abstract
Lung cancer remains the leading cause of cancer-related mortality worldwide, yet current screening criteria focusing on heavy smoking overlook a significant portion of at-risk individuals. To develop a robust predictive biomarker, we profiled circulating exosomal small non-coding RNAs (sncRNAs, including miRNAs, piRNAs, tiRNAs, and tRFs) in a prospective discovery cohort (UHCC; n = 202 smokers, with up to 16-year follow-up) and an independent validation cohort (CHTN; n = 186) including healthy individuals or patients with either malignant or benign tumor of the lung. A rigorous 5×5 nested cross-validation pipeline integrating differential expression, correlation pruning, and eight classifier comparisons identified a 31-sncRNA panel. Random forest was selected as the optimal model and yielded excellent discrimination in UHCC (AUC=0.97), with sensitivity 0.93 and specificity 1.00 at the Youden-optimized threshold. In the independent validation cohort, the signature demonstrated moderate performance, particularly in discriminating diagnosed cancer patients from healthy controls (AUC=0.73; AUPRC=0.85). Furthermore, the resulting risk score was strongly associated with incident lung cancer in the UHCC cohort, independent of demographic and smoking factors (multivariable OR=13.19, 95% CI 6.83-25.45), and predicted a shorter time-to-diagnosis in a Fine-Gray competing risks model (sHR=4.73, 95% CI 3.61-6.21) with a significant non-linear dose-response. Landmark and time-dependent analyses confirmed robust discrimination up to a decade pre-diagnosis, although precision was attenuated at longer intervals. At last, pathway analysis of the signature’s targets implicated key oncogenic pathways, including the PI3K-Akt, p53, MAPK, ErbB, and mTOR signaling pathways. This work establishes a panel of novel exosomal sncRNA signatures as a powerful risk prediction biomarker for lung cancer, enabling early risk stratification and creating a critical window for timely clinical intervention.
Disclosure
Z. Feng, None.. M. Nasu, None.. L. Higa, None.. I. M. Ibrahim, None.. L. L. Marchand, None.. Y. Deng, None.
Cited in
Control: 572 · Presentation Id: 10207 · Meeting 21436