Fig. 3
From: DeepMethyGene: a deep-learning model to predict gene expression using DNA methylations
Relationship between the number of methylation sites and predictive outcomes of geneEXPLORE and DeepMethyGene. (A) Illustrates that the number of methylation sites within a 1 Mb radius of the gene expression loci affects the predictive performance of both DeepMethyGene and geneEXPLORE. As observed, DeepMethyGene demonstrates superior performance compared to geneEXPLORE in regions with relatively lower methylation site density. (B) Illustrates the relationship between the differential predictive performance of DeepMethyGene and geneEXPLORE in terms of the R2 value for gene expression predictions and the number of methylation sites. Compare = R2(DeepMethyGene) − R2(geneEXPLORE). The results indicate that a lower number of methylation sites corresponds to a superior predictive performance of DeepMethyGene over geneEXPLORE