(a) (b)

Two models for the target DEG EVOVL5. The legends show the start base pair

d base pair of the DEG. The figures printed in the plots stand for the base pairs

ylation sites of the top-ranked DMSs. (a) The Lasso M2E model. (b) The RLR

l.

ne to methylation site association — E2M models

ܕ௞ stands for a differential methylation vector of a DMS, X

r a differential expression matrix of 1,250 DEGs and w stands for

of regression coefficients or model parameters which needs to be

d. There were in total 910 E2M regression models, each of which

ned as below,

ܕ௞ൌ݂ሺ܆, ܟሻ



methylation site of a DMS in such a regression model defined

as named as the target methylation site. In other words, each of

Ss was the target DMS in one regression model.

e 4.35 shows the density of the R-square measurements of four

dels. The four models worked very well, especially the random

d SVM models.

host gene of each target DMS was also identified. Based on

s (4.49) and (4.50), the gene order distance between the host gene

arget DMS and the top-ranked DEG was calculated. The densities

ne distances were estimated for the investigation of the skewness.

36 shows the distribution of the gene order distances between the

s of the target DMSs and the top-ranked DEGs of the four E2M