Abstract 4162: The identification of biomarker for the early diagnosis of IPMN associated PDAC based on serum protein profiling

Background: Most of pancreatic cancer patients are diagnosed at late stage due to the inability of early diagnosis for pancreatic ductal adenocarcinoma (PDAC), finally leading to a very high mortality. Therefore the identification of effective biomarkers for the early diagnosis of PDAC has a critical significance for improving the outcome of PDAC patients at early stage. Up to 15% of PDACs are thought to come from Mucinous pancreatic cysts, most of which belong to intraductal papillary mucinous neoplasm (IPMN). Consequently, screening and detecting high-risk IPMN patients at early stage by potential biomarkers can help a lot to prevent these patients from developing into invasive PDAC at late stage. Methods: We collected 38 normal serum sample as control and 243 IPMN serum samples with different progression stages including 83 adenomas, 73 borderlines, 40 carcinoma in situs and 47 invasive carcinomas. Then we performed antibody microarray experiment on all serum samples and implemented data analysis on serum protein profiling by R programming. After that, we divided the protein expression data into discovery set and validation set. With data pre-progressing, quality control, differential expression analysis and feature selection, we screened some protein panels for the construction of diagnostic model including logistic regression (LR) model and support vector machine (SVM) model based on discovery set. Finally, we further confirmed the diagnostic performance of these protein panels on IPMN associated PDAC based on validation set. Results: By data analysis, we finally constructed a stepwise diagnostic model of IPMN associated PDAC based on 11 protein panels with high sensitivity and specificity. For example, protein panel 1 containing 2 proteins can separate the normal and IPMN at all stages with sensitivity of 0.988 and specificity of 0.946 by SVM model; protein panel 2 containing 15 proteins can separate invasive carcinoma and non invasive carcinoma (adenoma, borderline and carcinoma in situ) with AUC of 0.99 by LR model; protein panel 3 containing 16 proteins can specifically separate invasive carcinoma from the normal with AUC of 0.978 by SVM model; protein panel 4 containing 8 proteins can specifically separate serum samples at early stages (adenoma and borderline) from the normal with AUC of 0.969 by SVM model, which is useful for the early diagnosis of IPMN associated PDAC; protein panel 5 containing 7 proteins can specifically separate serum samples at early-middle stage (adenoma, borderline and carcinoma in situ) from the normal with AUC of 0.987 by SVM model. In summary, this comprehensive diagnostic model can effectively accomplish IPMN serum samples classification for different tumor stages with high accuracy. Further validation on these protein panels with more IPMN serum samples by ELISA is in progress. Citation Format: Chaoyang Zhang, Mohamed Alhamdani, Andrea Bauer, Li Peng, Jorg Hoheisel. The identification of biomarker for the early diagnosis of IPMN associated PDAC based on serum protein profiling [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 4162.