In Clinical proteomics
Circulating biomarkers play important roles in diagnosis of malignant tumors. N-glycosylation is an important post-translation patter and obviously affect biological behaviors of malignant tumor cells. However, the role of N-glycosylation sites in early diagnosis of tumors still remains further investigation. In this study, plasma from 20 lung adenocarcinoma (LUAD), which were all classified as stage I, as well as 20 normal controls (NL) were labeled and screened by mass spectrometry (MS). Total 39 differential N-glycosylation sites were detected in LUAD, 17 were up-regulated and 22 were down-regulated. In all differential sites, ITGB3-680 showed highest potential in LUAD which showed 99.2% AUC, 95.0% SP and 95.0% SN. Besides, APOB-1523 (AUC: 89.0%, SP: 95.0%, SN: 70.0%), APOB-2982 (AUC: 86.8%, SP: 95.0%, SN: 45.0%) and LPAL2-101 (AUC: 81.1%, SP: 95.0%, SN: 47.4%) also acted as candidate biomarkers in LUAD. Combination analysis was then performed by random forest model, all samples were divided into training group (16 cases) and testing group (4 cases) and conducted by feature selection, machine learning, integrated model of classifier and model evaluation. And the results indicated that combination of differential sites could reach 100% AUC in both training and testing group. Taken together, our study revealed multiple N-glycosylation sites which could be applied as candidate biomarkers for early diagnosis diagnosis of LUAD.
Fang Kai, Long Qin, Liao Zhonghua, Zhang Chaoyu, Jiang Zhiqiang
2022-Nov-19
Circulating biomarkers, Early diagnosis, Lung adenocarcinoma, Machine learning, N-glycosylation
