Skip to main content
Fig. 1 | BMC Bioinformatics

Fig. 1

From: A convolutional neural network-based regression model to infer the epigenetic crosstalk responsible for CG methylation patterns

Fig. 1

Building epiNet models to predict CG methylation patterns based on other epigenetic features. a The epiNet model. Input FPKM values were transformed by one convolutional layer and one fully connected layer to predict the CG methylation levels of 1-kb bins (see “Methods” section). The shape of each layer is indicated within bracket. N: Number of input features. b The prediction of the CG methylation pattern of mouse FGOs based on varying numbers of input features. For each number of features (N = 1–8), the feature combination that showed the best correlation between the predicted and actual CG methylation patterns is shown. c A representative genome browser shot showing the predicted CG methylation patterns. The actual CG methylation, H3K36me3 and H3K4me3 patterns are shown for comparison. Genomic regions in which prediction was improved by the addition of the H3K4me3 data are highlighted in yellow. RefSeq genes are shown at the bottom. d Pairwise correlations between the actual data of all epigenetic features. Pearson correlation coefficients between the CG methylation pattern and the distributions of the respective features are indicated on the right

Back to article page