Result: CIRCA-CircularRNA for Cancer Active Immunotherapy: A Machine Learning Model to Predict Liver Cancer and Top Genes for Cancer Vaccine.

Circular RNAs (circRNAs) are long non-coding RNAs with excellent prognostic and diagnostic biomarker properties for many diseases including cancer. By using liver tissues of Hepatocellular Carcinoma (HCC) patient dataset, this study designed and tested a robust machine learning pipeline to predict HCC and circRNA targeted hub gene immunogenicity for immunotherapy. First, a publicly available circRNA microarray dataset was analyzed in Python for the top twelve deregulated circRNAs in tumor tissue compared to healthy tissue. Next, classification models were trained and tested on the circRNA data. microRNA (miRNA) and gene targets (mRNA) of deregulated circRNAs were predicted and top hub genes were found from gene interaction network analysis in Cytoscape. Finally, an immunogenicity predictor in Python was built with a T-cell epitope prediction framework. This study found: 1) hsa_circ_0005284 is strongly upregulated in tumor tissue and hsa_circRNA_ 089372 is strongly downregulated, 2) the Logistic Regression and Naive Bayes classification models most accurately predicted tumors from circRNA data, 3) the TMED10 and RAB1A hub genes were most immunogenic based on Python predictions. In conclusion, this project identifies hsa_circ_0005284 and hsa_circRNA_089372 as well as their linked immunogenic hub gene peptides as biological candidates for a liver cancer vaccine. [ABSTRACT FROM AUTHOR]