A Machine Learning Approach for Prediction of Domains of DELLA Proteins, a Key Component of Gibberellic Acid Signaling in Plants

Abstract

Background/Objective: For the annotation of large scale proteins, generally computational methods or tools are used. One of the drawbacks of these annotation tools is that they are not specific protein prediction programs. Methods/Analysis: In this study, we implement a machine-learning algorithm for fast and accurate prediction of DELLA proteins. Findings: We developed various modules by using conserved protein domains in DELLA proteins. To evaluate the modules classifiers like sequential minimum optimization, J48 decision tree, AD tree and logistic algorithms were used. By analyzing the results obtained from independent data set and cross-validation tests, maximum accuracy was achieved by logistic algorithm. The developed tool was tested with various inputs and it's showed that the algorithm developed in the study would be helpful in predicting plant DELLA domains. Applications: This tool will significantly contribute to deep level functional genome annotation and development of predictors.