Abstract

Abstract Background: Hepatocellular carcinoma (HCC) is the fastest rising cancer in the USA and is the fourth leading cause of cancer deaths globally. A critical unmet need to reduce high mortality associated with advanced HCC is the ability to identify high-risk individuals for cost-effective, targeted screening, early detection, and prevention strategies. Human genomics and animal models have revealed etiological patterns and multiple genes and signaling pathways such as TGF-β, WNT, VEGF to be associated with the initiation and progression of HCC. We used integrated functional approaches (combining bioinformatics analysis, in vivo mouse models, and in vitro biological and biochemical methods) and identified potential biomarkers for HCC. We examined and validated 10 functional biomarkers for HCC risk prediction. We further harnessed machine learning tools and Artificial Intelligence (AI)-based technology to validate manual analyses of Immunohistochemical (IHC) labeled slides across different centers. Methods: >280 liver tissue samples were collected from patients with HCC and cirrhosis. Immunohistochemistry was performed against 10 functional biomarkers (TGFBR1, TGFBR2, SPTBN1, HMGA2, RSPO3, ITIH4, EPCAM, PLK1, SIRT6, FANCD2). Artificial intelligence-based deep learning technology by AIFORIA was used for the unbiased, intensive, and rapid validation of IHC. Results: The most promising results were for TGFBR1 and TGFBR2, on >80 HCC and cirrhotic samples, IHC labeling revealed: • Both TGFBR1 (p<0.001) and TGFBR2 (p<0.02) showed considerably reduced expression levels in tumor tissue within HCC samples, compared to non-tumorous adjacent tissue and cirrhotic samples. • The results were validated with another set of 70 patient samples at two other centers. Results obtained at George Washington University (GW), University of Maryland (UMD), and the University of Hawaii (UH) for TGFBR2 were consistent. For TGFBR1, GW and UMD showed reproducible results while results at UH lacked statistical significance. • AI-based analysis for TGFBR2 (p<0.002) validated the results from manual analysis ensuring the consistency of results across different sites. • Of the other biomarkers, PLK1 also displayed potentially substantial alterations with increased labeling in tumor tissue as compared to non-tumor tissue. Conclusions: TGFBR2 is a promising functional predictive HCC tissue biomarker. • Tissue biomarkers are important as they provide spatial and contextual information. • Computerized analysis is a powerful tool to capture microscopic patterns in visual data that are not normally assessed in manual analysis such as the pattern of immune cells in cirrhosis and HCC. • AI adds consistency and confidence to results obtained at different sites. Citation Format: Sobia Zaidi, Kirty Shetty, Herbert Yu, Linda Wong, Shuyun Rao, Wilma Jogunoori, Richard Amdur, Shulin Li, Patricia latham, Bao-Ngoc Nguyen, Lopa Mishra. TGF-β receptors 1 and 2 are functional biomarkers that stratify risk of hepatocellular cancer (HCC). Artificial intelligence based validation at three centers [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 2544.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call