Model-based clustering of DNA methylation array data: a recursive-partitioning algorithm for high-dimensional data...

Area of Science:

• Genomics
• Epigenetics
• Bioinformatics

Background:

• Epigenetics studies heritable gene function changes without altering DNA sequence.
• Cytosine methylation is a key epigenetic mechanism for gene silencing, often implicated in cancer.
• DNA methylation arrays, like Illumina GoldenGate, analyze thousands of cancer-related genes, but scalable clustering remains a challenge.

Purpose of the Study:

• To develop a novel, scalable, and reliable clustering algorithm for DNA methylation array data.
• To address the limitations of existing clustering methods for high-throughput epigenomic data.

Main Methods:

• A model-based recursive-partitioning algorithm was developed to navigate clusters within a beta mixture model.
• Simulations were used to compare the proposed method against nonparametric and conventional mixture model approaches.

Main Results:

• The proposed algorithm demonstrated superior reliability compared to nonparametric methods and comparable reliability to conventional mixture models.
• The method offers improved computational efficiency over traditional mixture model approaches.
• Application to normal tissue samples revealed clusters associated with tissue type and age.

Conclusions:

• The recursively-partitioned mixture model provides an effective and computationally efficient solution for clustering DNA methylation data.
• This approach facilitates the identification of biologically relevant subgroups within epigenomic datasets.