Fig. 3From: Identifying subtype-specific associations between gene expression and DNA methylation profiles in breast cancerSubtype-specific Root Mean Squared Error from 5-fold cross validation. Each bar represents the average prediction error obtained from the proposed method (red), separate estimation that uses only the corresponding subtype data (yellow), and a single common estimation ignoring the subtype information (gray). Our proposed method shows significantly improved performance over the separate estimation approach, and slightly better or comparable performance over single common estimationBack to article page