Prediction of Protein Essentiality by the Support Vector Machine with Statistical Tests

Essential proteins include the minimum required set of proteins to support cell life. Identifying essential proteins is important for understanding the cellular processes of an organism. However, identifying essential proteins experimentally is extremely time-consuming and labor-intensive. Alternative methods must be developed to examine essential proteins. There were two goals in this study: identifying the important features and building learning machines for discriminating essential proteins. Data for Saccharomyces cerevisiae and Escherichia coli were used. We first collected information from a variety of sources. We next proposed a modified backward feature selection method and build support vector machines (SVM) predictors based on the selected features. To evaluate the performance, we conducted cross-validations for the originally imbalanced data set and the down-sampling balanced data set. The statistical tests were applied on the performance associated with obtained feature subsets to confirm their significance. In the first data set, our best values of F-measure and Matthews correlation coefficient (MCC) were 0.549 and 0.495 in the unbalanced experiments. For the balanced experiment, the best values of F-measure and MCC were 0.770 and 0.545, respectively. In the second data set, our best values of F-measure and MCC were 0.421 and 0.407 in the imbalanced experiments. For the balanced experiment, the best values of F-measure and MCC were 0.718 and 0.448, respectively. The experimental results show that our selected features are compact and the performance improved. Prediction can also be conducted by users at the following internet address: http://bio2.cse.nsysu.edu.tw/esspredict.aspx.