Predicting cervical cancer DNA methylation from genetic data using multivariate CMMP

Abstract Epigenetic modifications link the environment to gene expression and play a crucial role in tumour development. DNA methylation, in particular, is gaining attention in cancer research, including cervical cancer, the focus of this study. Public repositories such as The Cancer Genome Atlas (TCGA) provide extensive genetic profiles but comparatively limited epigenetic data. We propose a new method, called multivariate classified mixed model prediction (mvCMMP), a multivariate nested‐error regression framework for predicting DNA methylation from genetic data in cervical cancer. mvCMMP exploits dependencies among outcomes and class‐specific random effects associated with new observations. We show that mvCMMP improves prediction accuracy over competing methods, highlighting the benefits of borrowing strength across methylation markers and shared random effects.