Prediction of β-lactamase and its class by Chou's pseudo-amino acid composition and support vector machine

•We developed a two-level support vector machine based prediction system, PredLactamase, for β-lactamase protein and its class prediction.•At first level PredLactamase identifies β-lactamase from non-β-lactamase and at second level it classifies predicted lactamase into one of the 4 classes.•Performance of class B is higher among all classes, which indicates that metallo-lactamase is different from remaining three classes.•We also developed a user-friendly web-server and standalone, which can be downloaded from http://14.139.227.92/mkumar/predlactamase

β-Lactam class of antibiotics is used as major therapeutic agent against a number of pathogenic microbes. The widespread and indiscriminate use of antibiotics to treat bacterial infection has prompted evolution of several evading mechanisms from the lethal effect of antibiotics. β-Lactamases are endogenously produced enzyme that makes bacteria resistant against β-lactam antibiotics by cleaving the β-lactam ring. On the basis of primary structures, β-lactamase family of enzymes is divided into four classes namely A, B, C and D. Class B are metallo-enzymes while A, C and D does not need any metal in the enzyme catalysis. In the present study we developed a SVM based two level β-lactamases protein prediction method, which differentiate β-lactamases from non-β-lactamases at first level and then classify predicted β-lactamases into different classes at second level. We evaluated performance of different input vectors namely simple amino acid composition, Type-1 and Type-2 Chou's pseudo amino acid compositions. Comparative performances indicated that SVM model trained on Type-1 pseudo amino acid composition has the best performance. At first level we were able to classify β-lactamases from non-β-lactamases with 90.63% accuracy. At second level we found maximum accuracy of 61.82%, 89.09%, 70.91% and 70.91% of class A, class B, class C and class D, respectively. A web-server as well as standalone, PredLactamase, is also developed to make the method available to the scientific community, which can be accessed at http://14.139.227.92/mkumar/predlactamase.