Biomarker identification and cancer survival prediction using random spatial local best cat swarm and Bayesian optimized DNN

Abstract

Identifying cancer biomarkers is crucial for improving patient outcomes and reducing cancer-related deaths. This research proposes BioSurv, a framework for biomarker identification and cancer survival prediction, using machine learning and deep learning techniques. Multi-omics data from breast cancer (BRCA) and lung adenocarcinoma (LUAD), including mRNA, miRNA, CNV, and DNA methylation, are analyzed. The collected dataset is passed to statistical tests and the random spatial local best cat swarm optimization (RSLBCSO) algorithm for feature selection, followed by KEGG and survival analyses to identify prognostic markers. Thirteen BRCA and fifteen LUAD poor prognostic markers are identified. A Bayesian optimized deep neural network (DNN) is used for cancer survival prediction, achieving high accuracy of 90% and 91% for BRCA and LUAD, respectively.