Methylation Linear Discriminant Analysis (MLDA) for identifying differentially methylated CpG islands

Area of Science:

• Epigenetics
• Genomics
• Bioinformatics

Background:

• Promoter CpG island hypermethylation correlates with gene silencing and epigenetic maintenance.
• CpG island methylation plays a role in tumor development and chemotherapy resistance.
• Differential Methylation Hybridisation (DMH) is a technique for genome-wide DNA methylation analysis.

Purpose of the Study:

• To develop a novel algorithm, Methylation Linear Discriminant Analysis (MLDA), for analyzing DNA methylation microarray data.
• To account for the specific biological features of DNA methylation and data distribution in microarray analysis.
• To identify differentially methylated loci between cisplatin-sensitive and resistant ovarian cancer cell lines.

Main Methods:

• Developed the MLDA algorithm in R, utilizing linear regression models of non-normalized hybridisation data.
• Used log-transformed signal intensities of unmethylated controls as a reference for methylation status.
• Transformed signal intensities of digested DNA samples to determine the likelihood of locus methylation.

Main Results:

• MLDA identified 115 differentially methylated loci between cisplatin-sensitive and resistant ovarian cancer cell lines.
• 23 out of 26 validated loci were confirmed by Methylation Specific PCR and/or bisulphite pyrosequencing.
• The MLDA algorithm successfully identified differentially methylated loci using DMH data from CpG island microarrays.

Conclusions:

• MLDA offers advantages for analyzing CpG island microarray methylation data due to its clear definition of methylation status.
• MLDA utilizes DMH data without between-group normalization and is less influenced by cross-hybridisation.
• The MLDA algorithm effectively identified differentially methylated loci between sample classes analyzed by DMH.