betaclust: A Family of Beta Mixture Models for Clustering Beta-Valued DNA Methylation Data
Koyel Majumdar, Romina Silva, Antoinette Sabrina Perry, R. William G. Watson, Andréa Rau, Florence Jaffrézic, Thomas Brendan Murphy, Isobel Claire Gormley · 2022
A family of novel beta mixture models (BMMs) has been developed by Majumdar et al. (2022) to appositely model the beta-valued cytosine-guanine dinucleotide (CpG) sites, to objectively identify methylation state thresholds and to identify the differentially methylated CpG (DMC) sites using a model-based clustering approach. The family of beta mixture models employs different parameter constraints applicable to different study settings. The EM algorithm is used for parameter estimation, with a novel approximation during the M-step providing tractability and ensuring computational feasibility.