Exploring Multiple Biomarker Combination by Logistic Regression for Early Screening of Ovarian Cancer

The best marker combination for differentiating the ovarian cancer from benign is explored with the logistic regression. The serum samples from 81 patients with ovarian cancer and 216 patients with benign pelvic masses provided by 2 institutes were analyzed using Luminex assay test. The selection performance of the logistic regression was compared with three other methods such as t-test, genetic algorithm, and random forest. The evaluation of the four methods were performed also with three classification methods including logistic regression, linear discriminant analysis, and k-nearest neighbor method. The 4 marker combination from the logistic regression showed the best performance against the other selection methods in terms of the average accuracy.