Using IB1 Algorithm to Predict of Bacterial Toxins with an Improved Feature Extraction

Although bacterial toxins are a major cause of some diseases, the successful prediction of bacterial toxins directly from primary sequence is much benefited to further basic knowledge of cell biology or for medical research and application. In this paper, by using the concept of Chou’s pseudo amino acid composition, a new method was proposed to predict bacterial toxins by IB1 algorithm. The jackknife cross-validation is applied to test predictive capability of proposed method. The predictive result showed that the total prediction accuracy is 97.52% for bacterial toxins and non toxins, which is higher than previous methods. Furthermore, we also discriminated endotoxins and exotoxins by the proposed method, and obtained satisfactory result with a total prediction accuracy 94.67%. Our method may play a complementary role to other existing methods for predicting bacterial toxins