On the Prediction of non-CG DNA Methylation

DNA cytosine methylation is an epigenetic modification that has a critical role in gene regulation and genome stability. DNA methylation can be detected and measured using sequencing instruments after sodium bisulfite conversion, but experiments can be expensive for large eukaryotic genomes. Sequencing non-uniformity and mapping biases can leave parts of the genome with low or no coverage, thus hampering the ability of obtaining DNA methylation levels for all cytosines. To address these limitations, several computational methods have been proposed that can predict DNA methylation from the DNA sequence around the cytosine, or from the methylation level of nearby cytosines. Most of these methods are, however, entirely focused on CG methylation in humans and other mammals. In this work, we study for the first time the problem of predicting cytosine methylation for CG, CHG, and CHH contexts on five plant species, either from the DNA primary sequence around the cytosine or the methylation levels of neighboring cytosines. In this framework, we also study (1) the cross-species prediction problem, i.e., the classification performance when training on one species and testing on another species, and the (2) the cross-context prediction problem, i.e., the classification performance when training on one context and testing on another context (within the same species). Finally, we show that providing the classifier with gene annotation information allows our classifier to outperform the prediction accuracy of state-of-the-art methods.