Related Experiment Video
Updated: Aug 17, 2026
Ultrasound-Guided Orthotopic Implantation of Murine Pancreatic Ductal Adenocarcinoma
Published on: November 19, 2019
Identification of key genes in pancreatic ductal adenocarcinoma with biologically informed deep neural network
Ru Gao1, Ruichen Li1, Yichen Xing2
1School of Health Science and Engineering, University of Shanghai for Science and Technology, Shanghai, China.
Background:
Pancreatic ductal adenocarcinoma (PDAC) is a prevalent and lethal form of cancer. According to clinical trials results, immunotherapy has not been successful in PDAC so far, which may be due to the high heterogeneity of the tumour immune microenvironment (TIME) of PDAC patients. This study aimed to identify significant genes associated with the prognosis and immune microenvironment of PDAC using interpretive deep learning.
Methods:
The importance of genes was appraised by P-NET, a potent interpretive deep learning model based on a biologically informed neural network. In addition to differential expression and survival analyses, tumour immune dysfunction and exclusion (TIDE) and CellPhoneDB analyses were carried out to identify genes that exert an influence on the prognosis and immune microenvironment of PDAC. Finally, Library of Integrated Network-Based Cellular Signatures (LINCS) data and molecular docking techniques were employed to identify drugs that could potentially be repurposed for the treatment of PDAC.
Results:
Using P-NET, we identified 628 important genes from five independent PDAC cohorts. These genes were enriched in pathways related to cell proliferation and survival, including the PI3K-Akt signaling pathway, the Wnt signaling pathway, and the focal adhesion pathway. According to survival and cell communication analyses, we identified eight prognostic genes that may participate in tumour and immune cell crosstalk. Twenty compounds that could downregulate these genes were identified from LINCS data. Through molecular docking analysis, we found that ursolic acid (UA) and tanespimycin might target JAG1, MET, and PLAU. These three genes were highly expressed in PDAC subtypes associated with poor outcomes. TIDE analyses indicated that patients with high expression levels of these three genes exhibited resistance to immunotherapy.
Conclusions:
Our study demonstrated that JAG1, MET, and PLAU were significantly overexpressed and associated with poor outcomes in PDAC patients. More importantly, these genes are involved in the crosstalk between tumour and immune cells, which indicates that these genes may serve as novel targets for combination immunotherapy in the treatment of PDAC.
