EnMCB

EnMCB performs methylation correlated block partitioning and stacked-ensemble survival prediction from DNA methylation profiles, leveraging co-modification of contiguous cytosine-phosphorothioate-guanine (CpG) sites by methyltransferases or demethylases to identify epigenetic signatures for cancer progression and prognosis.

Key Features:

• Methylation Correlated Block Partitioning: Automatically partitions the genome into blocks of tightly co-methylated cytosine-phosphorothioate-guanine (CpG) sites to capture regional methylation patterns.
• Stacked Ensemble Modeling: Employs a stacked ensemble of machine learning models integrating Cox regression, support vector regression (SVR), mboost, and elastic-net to derive predictive methylation signatures from methylation correlated blocks.
• Survival Prediction: Applies ensemble-derived signatures to predict patient survival and disease progression using DNA methylation profiles from The Cancer Genome Atlas (TCGA).
• R/Bioconductor Implementation: Implemented as an R/Bioconductor package for analysis of DNA methylation data and model building.

Scientific Applications:

• Cancer prognosis and survival analysis: Generates methylation-based prognostic signatures and predicts patient survival in oncology cohorts.
• Epigenetic biomarker discovery: Identifies methylation correlated blocks as candidate diagnostic and prognostic epigenetic markers associated with cancer progression.
• TCGA cohort-level methylation analysis: Applies predictive modeling to DNA methylation profiles from The Cancer Genome Atlas (TCGA) for cohort-scale investigations.

Methodology:

Partitions the genome into methylation correlated blocks and applies a stacked ensemble combining Cox regression, support vector regression (SVR), mboost, and elastic-net to build predictive signatures from DNA methylation profiles such as those from TCGA.