Quantitative identification of differentially methylated loci based on relative entropy for matched case-control data

As an important epigenetic modification, DNA methylation plays a critical role in regulating multiple biological processes and diseases. Many efforts have been devoted to identifying differentially methylated loci (DML) between cases and controls.

However, most present methods are statistical and are limited in handling methylation data with characteristics of high heterogeneity and non-normal distribution. Here, a quantitative method, quantitative DML (QDML), based on modified relative entropy is introduced to face these challenges, which can identify DML, hypermethylated loci and hypomethylated loci simultaneously. QDML, compared with some statistical methods, does not require a presupposed distribution of methylation data. Furthermore, QDML is more powerful in handling highly heterogeneous data, owing to the difference in sensitivity on every matched sample pair in case-control groups rather than the overall difference of all samples.

Simulation studies and real-data application show that QDML has a higher accuracy and a lower false-positive rate when identifying DML than statistical methods.

QDML is developed to identify DML based on relative entropy that can quantify the difference in methylation status between cases and controls. Its applications are not limited to methylation data and can be extended to other case-control studies.