Predicting methylation status of human DNA sequences by pseudo-trinucleotide composition

Abstract

DNA methylation plays a key role in the regulation of gene expression. The most common type of DNA modification consists of the methylation of cytosine in the CpG dinucleotide. The detections of DNA methylation have been determined mostly by experimental methods; however, these methods were time-consuming, expensive, and difficult to meet the requirements of modern large-scale sequencing technology. Accordingly, it is necessary to develop automatic and reliable prediction methods for DNA methylation. In this study, the pseudo-trinucleotide composition was proposed, and a novel method was developed by support vector machine (SVM) with the pseudo-trinucleotide composition as input parameter to represent DNA sequence for DNA methylation prediction. The model was evaluated on two datasets, including a dataset of Rollins (dataset_1) and a dataset collected healthy human records from the MethDB database (dataset_2). For dataset_1, the Matthews correlation coefficient (MCC) and accuracy (ACC) by jackknife validation were 0.8051 and 0.6098, respectively. For dataset_2, the MCC and ACC were 0.8500 and 0.7203, respectively. The good prediction results reveal that the pseudo-trinucleotide composition is an effective representation method for DNA sequence and plays a very important role in the prediction of DNA function.