A Cholangiocarcinoma Prediction Model Based on Random Forest and Artificial Neural Network Algorithm

• Published in: Journal of the College of Physicians and Surgeons--Pakistan Vol. 33; no. 5; pp. 578 - 586
• Main Authors: Liao, Jianhua, Meng, Chunyan, Liu, Baoqing, Zheng, Mengxia, Qin, Jun
• Format: Journal Article
• Language: English
• Published: Pakistan College of Physicians and Surgeons Pakistan 01.05.2023
• Subjects: Algorithms; Analysis; Bile Duct Neoplasms - diagnosis; Bile Duct Neoplasms - genetics; Bile Ducts, Intrahepatic - pathology; Cholangiocarcinoma - diagnosis; Cholangiocarcinoma - genetics; Gene expression; Humans; Machine learning; Neural networks; Neural Networks, Computer; Random Forest; Tumor Microenvironment; China
• ISSN: 1022-386X; 1681-7168; 1681-7168
• DOI: 10.29271/jcpsp.2023.05.578

To construct a prognostic model using artificial neural network (ANN) approach, providing an idea for the prediction and diagnosis of cholangiocarcinoma (CCA). Experimental study. Place and Duration of the Study: Department of General Surgery, Zhenjiang Hospital, Zhenjiang Province, China, between January and March 2022. Available datasets were obtained from the Gene Expression Omnibus (GEO) database to construct the train cohort and the test cohort of CCA, and screened out the differentially expressed genes (DEGs) of CCA. Next, an ANN model for CCA diagnosis was constructed based on the scores of the DEGs and evaluated its accuracy and efficiency using ROC curves. Finally, the immune infiltration and the function of extracellular matrix (ECM) protein SPACRL1 were analysed to reveal the characteristic alterations in CCA. This analysis revealed 166 DEGs, mainly concentrated in the ECM organisation, neutrophil activation and other pathways. Then a set of 17 CCA disease signature genes scores were obtained to build an ANN prediction model and the ROC curve was plotted. The AUC in the train group (0.980) indicated that the accuracy of the diagnosis model is extremely high. Finally, there was a significant increase of B cells naïve (p=0.025), tregs (p=0.004), and macrophages M1 (p<0.001) in the tumour-microenvironment of CCA, while SPARCL1 was a protective factor on disease-specific survival (DSS) in CCA (p=0.009). This study has developed an accurate prediction model for CCA diagnosis, and identified SPARCL1 as pivotal factor in CCA by modulating the tumour immune-microenvironment. Cholangiocarcinoma, Artificial neural network, Immune microenvironment, Bioinformatics, Prognosis model, SPARCL1.