PDAC Prediction with a Urine Biomarker Penal
Yujie (Janet) He
Yujie (Janet) He
OVERVIEW
Pancreatic ductal adenocarcinoma (PDAC) is one of the most lethal diseases, even within the spectrum of cancers. Only around 9% of patients can survive for more than 5 years once diagnosed due to its asymptomatic property in its infancy when surgery can be curative.
The late diagnosis of PDAC is primarily due to the lack in reliable biomarkers for detection in its initial stage. Biomedical scientists have been making efforts to achieve a higher surviving rate with possible non-invasive early-detection methods in the past few years.
Pancreas anatomy. Source: American Cancer Society
In 2020, Crnogorac-Jurcevic et al. [1] described a promising biomarker panel in urine. In this panel, LYVE1 (lymphatic vessel endothelial hyaluronan receptor 1) is a receptor that binds to both soluble and immobilized hyaluronan. REG1A and REG1B (regenerating family member 1 alpha and beta) belong to a family of REG glycoproteins, a group of calcium-dependent proteins expressed in pancreatic acinar cells. Both of them act as autocrine and paracrine growth factors and have been described in patients with pancreatic diseases and during pancreatic islet regeneration. TFF1 (trefoil factor 1) belongs to a family of gastrointestinal secretory peptides that interact with mucins and are expressed at increased levels during reconstitution and repair of mucosal injury. They protect epithelial cells from apoptotic death and increase their motility, but also play similar pivotal roles in cancer cells, and are thus involved in the development and progression of various cancer types.
Despite that blood has traditionally been the main source of these markers, urine becomes a competitive alternative biological fluid due to its non-invasive sampling, large volume collection and narrower dynamic range because of the less complicated components. Statistical models are great tools to assess the capability of diagnosing pancreatic cancer with urine as well as comparing the weights of various biomarker levels in detection.
In this project, I assessed the potential of using this specific urine biomarker panel for pancreatic cancer diagnostics by building a statistic model between the levels of the suspicious protein biomarkers and the diagnosis.
With traditional statistical methods, A logistic regression model was built and the parameters were estimated in different ways. Novel machine learning (ML) models were also used and assessed in parallel. Different models were benchmarked and compared through their accuracies for prediction.
[1] S. Debernardi, H. O’Brien, A. S. Algahmdi, N. Malats, G. D. Stewart, M. Pljeˇsa- Ercegovac, E. Costello, W. Greenhalf, A. Saad, R. Roberts, et al. A combination of urinary biomarker panel and pancrisk score for earlier detection of pancreatic cancer: A case–control study. PLoS medicine, 17(12):e1003489, 2020.