Random forest model and GO enrichment. a, Schematic plot of random forest; b, GO terms enriched by the matched proteins of differential peptide probes.
ARTICLE ABSTRACTEarly diagnosis is critical to lung adenocarcinoma patients’ survival but faces inadequacies in convenient early detection.
We applied a comprehensive microarray of 130,000 peptides to detect “autoantibody signature” that is autoantibodies binding to mimotopes for early detection of stage 0–I LUAD. Plasma samples were collected from 147 early-stage lung adenocarcinoma (Early-LUAD), 108 benign lung disease (BLD), and 122 normal healthy controls (NHC). Clinical characteristics, low-dose CT (LDCT), and laboratory tests were incorporated into correlation analysis.
We identified 143 and 133 autoantibody signatures, distinguishing Early-LUAD from NHC/BLD in the discovery cohort. Autoantibody signatures significantly correlated with age, stage, tumor size, basophil count, and IgM level (P < 0.05). The random forest models based on differential autoantibody signatures displayed AUC of 0.92 and 0.87 to discern Early-LUAD from NHC/BLD in the validation cohort, respectively. Compared with LDCT, combining autoantibody signature and LDCT improved the positive predictive value from 50% to 78.33% (P = 0.049). In addition, autoantibody signatures displayed higher sensitivity of 72.4% to 81.0% compared with the combinational tumor markers (cyfra21.1, NSE, SCC, ProGRP) with a sensitivity of 22.4% (P = 0.000). Proteins matched by differential peptides were enriched in cancer-related PI3K/Akt, MAPK, and Wnt pathways. Overlaps between matched epitopes and autoantibody signatures illustrated the underlying engagement of autoantibodies in immune recognition.
Collectively, autoantibody signatures identified by a high-throughput peptide microarray have the potential to detect Early-LUAD, which could assist LDCT to better diagnose Early-LUAD.
Novel sensitive autoantibody signatures can adjuvant LDCT to better diagnose LUAD at very early stage.