Identifying genes indispensable for an organism‘s life and theircharacteristics is one of the central questions in currentbiological research, and hence it would be helpful to developcomputational approaches towards the prediction of essentialgenes. The performance of a predictor is usually measured bythe area under the receiver operating characteristic curve(AUC). We propose a novel method by implementing geneticalgorithms to maximize the partial AUC that is restricted to aspecific interval of lower false positive rate (FPR), the regionrelevant to follow-up experimental validation. Our predictoruses various features based on sequence information, proteinproteininteraction network topology, and gene expressionprofiles. A feature selection wrapper was developed toalleviate the over-fitting problem and to weigh each feature’srelevance to prediction. We evaluated our method using theproteome of budding yeast. Our implementation of geneticalgorithms maximizing the partial AUC below 0.05 or 0.10 ofFPR outperformed other popular classification methods. [BMBReports 2013; 46(1): 41-46]