Ovarian Cancer Proteome Analysis and Biomarker Discovery Using Machine Learning

Ovarian cancer’s significance lies in its status as the most lethal gynecological cancer. Enhancing early detection and raising awareness are crucial factors that can enhance survival rates and the well-being of those impacted by the disease. Machine learning has a profound impact on diagnosing and detecting several types of cancers at an early stage. The ElasticNet algorithm is used to extract noteworthy features from diverse protein categories to identify biomarkers specific to each type. To validate the findings, the support vector machine algorithm was employed to classify unseen samples according to the identified biomarkers. In this study, the CPTAC -Ovarian cancer dataset was utilized, consisting of three distinct Omics-data types: N-glycoproteome, Phosphoproteome, and Proteome. Initially, SVM training was conducted for each dataset separately, followed by testing. The obtained results demonstrated high accuracy rates of 100% for N-glycoproteome, 100% for Phosphoproteome, and 96% for Proteome datasets. Remarkably, when ElasticNet and SVM were utilized, the experimental results demonstrated 100% accuracy for all three types. Additionally, the processing time was reduced by more than 50% when classifying based on only 20 selected features.