An SVM-based algorithm for the prediction and classification of enzymes involved in antibiotic biosynthetic pathways in plant growth promoting Pseudomonas species

In this study, a tool has been developed for the prediction of enzymes involved in antibiotic biosynthetic pathways(2,4-diacetylphloroglucinol, phenazine, pyoluteorin and pyrrolnitrin) in plant growth promoting Pseudomonas species on the basis of amino acid and dipeptide composition by using the Support Vector Machines (SVM). The performance of the system was achieved by using a training set consisting of 330 non-redundant set of positively labeled enzymes involved in antibiotic biosynthetic pathway in Pseudomonas spp. and 309 non-redundant set of negatively labeled sequences from other organisms obtained from NCBI. First we developed a support vector machine based module using amino acid and dipeptide composition and achieved an overall accuracy of 87.00% and 91.00% respectively. Then, another SVM module was developed based on dipeptide composition for classifying the predicted enzymes into four main classes with accuracy 95%, 80%, and 75% 95% for 2,4-diacetylphloroglucinol, phenazine, pyoluteorin and pyrrolnitrin respectively. Based on the above method, a web server has been set up at http://210.212.229.59:8080/Prediction/home.jsp.