Prediction of Viral Replication Origins Using Least-Squares Support Vector Machines

According to existing research, the replication origins (ORI) of certain viruses are located in DNA regions especially rich of repetitive sequences. Based on this observation, statistical methods for identifying likely sites of ORIs have been developed. In this presentation, some new relevant features, not associated with repetitive structures, are introduced. Then, Support Vector Machines (SVMs) are utilized to pinpoint the locations of replication origins in herpesviruses and caudoviruses. The SVMs achieved sensitivity and positive predictive value superior or at least equal to those provided by the previous sources. Also, as a byproduct of this process, the relative significance of each feature was estimated. Finally, a simple approach to obtain predictions, by combining the results of different techniques, is described.